本次给大家整理的是《International Journal of Applied Earth Observation and Geoinformation》杂志2024年08月第132期的论文的题目和摘要，一共包括88篇SCI论文！由于论文过多，我们将通过两篇文章进行介绍，本篇文章介绍第45--第88篇论文!
论文45
Quantitative inversion of soil trace elements from spectroscopic effects across multiple crop growth periods
基于光谱效应的土壤微量元素定量反演：涵盖多个作物生长周期
【摘要】
Human health is directly impacted by trace components found in soil, for which hyperspectral remote sensing technology offers a novel way to rapidly assess dynamic soil environmental quality. In most studies, the premise of quantitative inversion on soil elements is that the known target elements will cause growth stress. However, such stress is unusual in non-polluted areas. Consequently, broad-area soil trace element monitoring in non-contaminated areas remains challenging. Spectral inversion of plant material has a longer time window and is more sensitive to elemental changes than soil spectral inversion (bare soil period). In this work, we conducted a wheat pot experiment with concentration gradients of four trace elements (Fe, B, Mo, and Zn), analyzed leaf (jointing stage) and spike (maturity period) samples using spectral measurements and chemical tests, and filtered the characteristic positions and inverse modeling using the spectra of the element-accumulating preference organs in plant canopies. The canopy aggregation sites differed for each element, and both simple linear regression (SLR) and multiple linear regression (MLR) models based on the spectra of canopy aggregation sites achieved high accuracy. The results of this study enable the construction of an inverse model of plant spectra-plant element content-elements in soil, which can serve as a reference for soil monitoring and assessment in typical crop cover areas.
【摘要翻译】
人类健康直接受到土壤中微量成分的影响， hyperspectral 遥感技术提供了一种快速评估动态土壤环境质量的新方法。在大多数研究中，土壤元素的定量反演前提是已知的目标元素会引起植物生长压力。然而，这种压力在未污染地区不常见。因此，在非污染区域进行广泛土壤微量元素监测仍然面临挑战。植物材料的光谱反演具有更长的时间窗口，并且对元素变化比土壤光谱反演（裸土时期）更敏感。在本研究中，我们进行了一个小麦盆栽实验，涉及四种微量元素（Fe、B、Mo 和 Zn）的浓度梯度，使用光谱测量和化学测试分析了叶片（分蘖期）和穗（成熟期）样本，并利用植物冠层中元素积累偏好的光谱过滤特征位置和反演建模。不同元素的冠层聚集位置各异，基于冠层聚集位置光谱的简单线性回归（SLR）和多元线性回归（MLR）模型都实现了高精度。本研究的结果使得植物光谱-植物元素含量-土壤元素的反演模型构建成为可能，这可以作为典型作物覆盖区域土壤监测和评估的参考。
【doi】
https://doi.org/10.1016/j.jag.2024.104059
【作者信息】
Daming Wang, 天津中心，中国地质调查局，天津，300170，中国；华北地球科学创新中心，天津，300170，中国
Shawn W. Laffan, 新南威尔士大学生物、地球与环境科学学院，悉尼，新南威尔士州 2052，澳大利亚
Jing Zhang, 中国地质调查局天津中心，天津 300170，中国；华北地质科学创新中心，天津 300170，中国
Surong Zhang, 中国地质调查局天津中心，天津 300170，中国；华北地质科学创新中心，天津 300170，中国
Xusheng Li，中国地质调查局天津中心，天津 300170，中国；华北地质科学创新中心，天津 300170，中国
论文46
First evidence of a geodetic anomaly in the Campi Flegrei caldera (Italy) ground deformation pattern revealed by DInSAR and GNSS measurements during the 2021–2023 escalating unrest phase
2021–2023年Campi Flegrei火山（意大利）地面变形模式中首次揭示的地质异常证据：DInSAR和GNSS测量结果
【摘要】
Campi Flegrei caldera is an Italian high-risk volcano experiencing a progressively more intense long-term uplift, accompanied by increasing seismicity and geochemical emissions over the last two decades. Ground deformation shows an axisymmetric bell-shaped pattern, with a maximum uplift of about 120 cm, from 2005, in the caldera central area. We analyzed Sentinel-1 and COSMO-SkyMed Multi-Temporal DInSAR measurements and GNSS data to reveal and investigate a geodetic anomaly that has clearly manifested since 2021, locally deviating from the typical bell-shaped deformation pattern. This anomaly is located east of Pozzuoli town, in the Mt. Olibano–Accademia area, covers an area of about 1.3 km2 and shows, in comparison to surrounding areas, a maximum uplift deficit of about 9 cm between 2021 and 2023. To investigate the anomaly causes, we analyzed the caldera seismicity and inverted the DInSAR data to determine the primary source of the ground deformation pattern, which is consistent with a penny-shaped source located approximately 3800 m beneath the Pozzuoli town, with a radius of about 1200 m. We also found that the time evolution of the uplift deficit in the geodetic anomaly area correlates well with the earthquake occurrence, with the greater magnitude events clustering in this area. These considerations suggest the geodetic anomaly is a local response to the tensile stress regime produced by the inflating primary deformation source. This phenomenon can be influenced by the Mt. Olibano–Accademia lava domes lithological heterogeneities that may induce a localized reaction to ground deformation during the inflationary phase. Our interpretation aligns with the concentration of earthquakes and hydrothermal fluid emissions in this area, indicating the presence of faults, fractures, and fluid circulation. Accordingly, the geodetic anomaly area represents a zone of crustal weakness that requires careful monitoring and study.
【摘要翻译】
Campi Flegrei火山口是意大利的一座高风险火山，近年来经历了越来越强烈的长期隆升，同时伴随有不断增加的地震活动和气体排放。地面变形显示出一个轴对称的钟形模式，自2005年以来，在火山口中央区域的最大隆升约为120厘米。我们分析了Sentinel-1和COSMO-SkyMed多时相DInSAR测量数据以及GNSS数据，以揭示和研究自2021年以来明显出现的地质异常，该异常局部偏离了典型的钟形变形模式。这个异常位于波佐利市东部，即Mt. Olibano–Accademia区域，覆盖约1.3平方公里，与周围地区相比，2021年至2023年间的最大隆升缺口约为9厘米。为了研究异常的原因，我们分析了火山口的地震活动，并反演了DInSAR数据，以确定地面变形模式的主要来源，该来源与一个位于波佐利市下方约3800米处的半径约1200米的圆盘状源一致。我们还发现，地质异常区域的隆升缺口的时间演变与地震发生情况有很好的相关性，大震事件在该区域集中。这些考虑表明，该地质异常是对膨胀的主要变形源产生的拉张应力状态的局部响应。该现象可能受到Mt. Olibano–Accademia熔岩穹丘的岩石异质性的影响，这可能在膨胀阶段引起局部的地面变形反应。我们的解释与该区域地震和热液流体排放的集中情况一致，表明该区域存在断层、裂缝和流体循环。因此，地质异常区域代表了一个需要仔细监测和研究的地壳薄弱区。
【doi】
https://doi.org/10.1016/j.jag.2024.104060
【作者信息】
Flora Giudicepietro, 意大利国家地球物理和火山研究所，维苏威火山观测台，维苏威大道 328 号，80134 那不勒斯，意大利；意大利国家研究委员会，环境电磁探测研究所，那不勒斯-米兰，意大利
Francesco Casu, 意大利国家研究委员会，环境电磁探测研究所，那不勒斯-米兰，意大利
Manuela Bonano, 意大利国家研究委员会，环境电磁探测研究所，那不勒斯-米兰，意大利
Claudio De Luca, 意大利国家研究委员会，环境电磁探测研究所，那不勒斯-米兰，意大利
Prospero De Martino, 意大利国家地球物理和火山学研究所，维苏威火山观测站，Diocleziano街328号，80134 那不勒斯，意大利；意大利国家研究委员会，环境电磁探测研究所，那不勒斯-米兰，意大利
Federico Di Traglia, 意大利国家地球物理与火山学研究所，维苏威火山观测站，Diocleziano街328号，80134 那不勒斯，意大利；意大利国家研究委员会，环境电磁探测研究所，那不勒斯-米兰，意大利
Mauro Antonio Di Vito, 意大利国家地球物理与火山学研究所，维苏威火山观测站，Diocleziano街328号，80134 那不勒斯，意大利
Giovanni Macedonio, 意大利国家地球物理与火山学研究所，维苏威火山观测站，Diocleziano街328号，80134 那不勒斯，意大利；意大利国家研究委员会，环境电磁探测研究所，那不勒斯-米兰，意大利
Michele Manunta, 意大利国家研究委员会，环境电磁探测研究所，那不勒斯-米兰，意大利
Fernando Monterroso, 意大利国家研究委员会，环境电磁探测研究所，那不勒斯-米兰，意大利
Pasquale Striano, 意大利国家研究委员会，环境电磁探测研究所，那不勒斯-米兰，意大利
Riccardo Lanari，意大利国家研究委员会，环境电磁探测研究所，那不勒斯-米兰，意大利
论文47
A novel Slepian approach for determining mass-term sea level from GRACE over the South China Sea
一种新颖的Slepian方法用于从GRACE数据确定南中国海的质量项海平面
【摘要】
Treatment of land-to-ocean leakage is crucial in modeling the mass-term sea level (MSL) using the Gravity Recovery and Climate Experiment (GRACE) satellite gravity measurements. In this study, we utilized the Spherical Slepian functions (SSFs) to determine MSL variations in the South China Sea. A sensitivity simulation in terms of trend, annual amplitude, and phase revealed that the SSF solution with a 1° coastal buffer zone provides a better land-to-ocean leakage correction than traditionally used Spherical Harmonics (SH). It was also found that an additional smoothing procedure for SSF with low concentrating energy could significantly reduce the high-frequency noise in GRACE (e.g., the north–south strips). The spatiotemporal characteristics of the true MSL were further compared among GRACE SH and Mascon solutions, model-predicted ocean bottom pressure, and steric-corrected altimetric data (i.e., satellite-altimetric sea level minus steric effect). Results revealed that despite the SSF-inverted regional MSL solution being generally similar to other results, this technique notably recovered the realistic magnitude of detailed features. The apparent MSL signal was, for instance, precisely recovered in the central part of the Gulf of Thailand and the Sunda Shelf by the SSF, while the SH significantly underestimated it due to smoothing. In addition to the seasonality, the interannual signal decomposed from the SSF-inverted residual MSL (by removing the trend and periodic signals) was the strongest and well cross-correlated with the Southern Oscillation Index. With more detailed spatial patterns revealed by MSL from GRACE SSF, our findings demonstrate that SSF is more suitable for regional-scale studies.
【摘要翻译】
在利用重力恢复与气候实验（GRACE）卫星重力测量数据建模海平面质量项（MSL）时，处理陆地到海洋的泄漏至关重要。在本研究中，我们使用了球面斯莱皮安函数（SSF）来确定南中国海的MSL变化。趋势、年振幅和相位的敏感性模拟结果表明，具有1°沿海缓冲区的SSF解决方案提供了比传统使用的球面调和函数（SH）更好的陆地到海洋泄漏修正。研究还发现，对SSF进行额外的平滑处理可以显著减少GRACE数据中的高频噪声（例如南北条纹）。我们进一步比较了真实MSL的时空特征，包括GRACE SH和Mascon解决方案、模型预测的海底压力以及修正了 steric 效应的高程数据（即卫星高度计海平面减去 steric 效应）。结果显示，尽管SSF反演的区域MSL解与其他结果通常相似，但该技术显著恢复了详细特征的真实幅度。例如，SSF精确恢复了泰国湾中部和巽他架的明显MSL信号，而SH由于平滑处理显著低估了该信号。此外，SSF反演的残余MSL（去除趋势和周期信号后）分解出的年际信号最强，并与南方涛动指数良好相关。通过GRACE SSF揭示的更详细的空间模式，我们的研究表明，SSF更适合于区域尺度的研究。
【doi】
https://doi.org/10.1016/j.jag.2024.104065
【作者信息】
Zhongtian Ma, 香港理工大学土地测量与地理信息学系，中国香港
Hok Sum Fok, 武汉大学测绘与遥感学院，中国武汉；武汉大学地空间环境与大地测量教育部重点实验室，中国武汉；湖北罗家实验室，中国武汉
Robert Tenzer, 香港理工大学土地测量与地理资讯系，中国香港
Jianli Chen，香港理工大学土地测量与地理资讯系，中国香港；香港理工大学土地与空间研究所，中国香港
论文48
Enhanced cotton chlorophyll content estimation with UAV multispectral and LiDAR constrained SCOPE model
基于无人机多光谱和LiDAR约束SCOPE模型的棉花叶绿素含量增强估计
【摘要】
Accurate and non-destructive estimation of leaf chlorophyll content (LCC) is crucial for optimizing cotton production. This study enhances the SCOPE model by integrating unmanned aerial vehicle (UAV)-derived multispectral data with leaf area index (LAI) from LiDAR data, significantly improving precision of LCC estimation, particularly during crucial growth stages of cotton. We construct and analyze three cost functions: COST1, which relies solely on spectral data; COST2, which incorporates direct LAI inputs; and COST3, which adjusts for LAI measurement uncertainties by combining spectral term with an error term representing the squared relative error between measured and model-estimated LAI. Our findings indicate that while COST1 establishes a baseline, COST2 and COST3 provide more accurate LCC estimations. COST3, validated against theoretical data, field-measured cotton datasets, and an additional maize dataset, proves most robust, maintaining consistent accuracy across all growth stages especially when considering input data uncertainties. This highlights the importance of integrating appropriate forms of LAI in cost functions to refine LCC estimation. Future research should focus on improving data acquisition quality and developing more advanced cost functions to advance LCC estimation further.
【摘要翻译】
准确且无损地估计叶片氯ophyll含量（LCC）对优化棉花生产至关重要。本研究通过将无人机（UAV）获取的多光谱数据与LiDAR数据的叶面积指数（LAI）整合，增强了SCOPE模型，从而显著提高了LCC估计的精度，特别是在棉花生长的关键阶段。我们构建并分析了三种成本函数：COST1，仅依赖光谱数据；COST2，包含直接的LAI输入；以及COST3，通过将光谱项与表示测量与模型估计LAI之间平方相对误差的误差项结合，调整LAI测量的不确定性。研究发现，虽然COST1建立了一个基准，但COST2和COST3提供了更准确的LCC估计。COST3在理论数据、现场测量的棉花数据集和额外的玉米数据集上的验证中表现最为稳健，特别是在考虑输入数据不确定性时，保持了一致的准确性。这突显了在成本函数中整合适当的LAI形式以改进LCC估计的重要性。未来的研究应集中于提高数据采集质量并开发更先进的成本函数，以进一步推进LCC估计。
【doi】
https://doi.org/10.1016/j.jag.2024.104052
【作者信息】
Puchen Yan, 国家农业水资源高效利用重点实验室，中国，北京；国家绿洲农业高效用水野外科学观测研究站，中国，武威 733009；中国农业大学农业水研究中心，中国，北京 100083
Yangming Feng, 国家农业水资源高效利用重点实验室，北京，中国；国家绿洲农业高效用水野外科学观测研究站，武威 733009，中国；中国农业大学农业水研究中心，北京 100083，中国
Qisheng Han, 中国农业科学院农田灌溉研究所，河南省新乡市 453003，中国
Zongguang Hu, 中国农业水资源高效利用国家重点实验室，北京，中国；国家绿洲农业高效用水野外科学观测研究站，武威 733009，中国；中国农业大学农业水资源研究中心，北京 100083，中国
Xi Huang, 北京市农业水资源高效利用国家重点实验室，中国；国家绿洲农业高效用水科学观测与研究站，武威 733009，中国；中国农业大学农业水资源研究中心，北京 100083，中国
Kaikai Su, 北京市农业水资源高效利用国家重点实验室，中国；国家绿洲农业高效用水科学观测与研究站，武威 733009，中国；中国农业大学农业水资源研究中心，北京 100083，中国
Shaozhong Kang，北京市农业水资源高效利用国家重点实验室，中国；国家绿洲农业高效用水科学观测与研究站，武威 733009，中国；中国农业大学农业水资源研究中心，北京 100083，中国
论文49
A spatio-temporal unmixing with heterogeneity model for the identification of remotely sensed MODIS aerosols: Exemplified by the case of Africa
一种用于识别遥感MODIS气溶胶的时空解混合与异质性模型：以非洲为例
【摘要】
Aerosols are crucial constituents of the atmosphere, with significant impacts on air quality. Aerosol optical depth (AOD) is critical in assessing solar resources and modeling sky radiance. However, comprehensive aerosol studies at a continental scale are limited, and existing methodologies need to consider spatial characteristics. This study develops a spatio-temporal unmixing with heterogeneity (STUH) model to evaluate spatial patterns and temporal trends of atmospheric aerosols across the African continent. The spatio-temporal AOD data cube, comprising monthly averaged MODIS-derived AOD data from 2001 to 2015, was decomposed using spatially non-negative matrix variabilization to explore the spatial determinants and the impacts of their interactions to AOD using a geographically optimal zones-based heterogeneity (GOZH) model. Our findings reveal an increasing trend of aerosol levels across Africa in the past 15 years, combined with the spatio-temporal AOD pattern explained by five abundance variables. We find that in different regions across Africa, the impact of natural variables on AOD was 1.56 to 3.01 times the impact of human variables, with significant spatial variations. These results are essential for understanding the climatic implications of atmospheric aerosols in Africa.
【摘要翻译】
气溶胶是大气的重要成分，对空气质量具有显著影响。气溶胶光学厚度（AOD）在评估太阳资源和模拟天空辐射中至关重要。然而，现有的大规模气溶胶研究有限，现有方法需要考虑空间特征。本研究开发了一个带有异质性的时空解混合（STUH）模型，用于评估非洲大陆大气气溶胶的空间模式和时间趋势。通过使用2001年至2015年每月平均的MODIS气溶胶光学厚度数据构建的时空AOD数据立方体，采用空间非负矩阵变分化方法进行分解，以探讨空间决定因素及其交互作用对AOD的影响，并结合地理最优区域异质性（GOZH）模型进行分析。我们的研究发现，过去15年中，非洲气溶胶水平呈上升趋势，且时空AOD模式由五个丰度变量解释。在非洲不同地区，自然变量对AOD的影响是人为变量影响的1.56到3.01倍，且存在显著的空间差异。这些结果对理解气溶胶对非洲气候的影响具有重要意义。
【doi】
https://doi.org/10.1016/j.jag.2024.104068
【作者信息】
Longshan Yang, 贵州大学矿业学院，贵阳，中国
Peng Luo, 慕尼黑工业大学航空航天与测绘系制图学教席，慕尼黑，德国；麻省理工学院感知城市实验室，剑桥，美国
Zehua Zhang, 科廷大学设计与建筑环境学院，珀斯，澳大利亚
Yongze Song, 科廷大学设计与建筑环境学院，珀斯，澳大利亚
Kai Ren, 科廷大学设计与建筑环境学院，澳大利亚珀斯；西北农林科技大学自然资源与环境学院，中国陕西杨凌 712100
Ce Zhang, 布里斯托大学地理科学学院，英国布里斯托 BS8 1SS
Joseph Awange, 地球与行星科学学院，空间科学学科，科廷大学，澳大利亚珀斯
Peter M. Atkinson, 兰卡斯特环境中心，兰卡斯特大学，兰卡斯特 LA1 4YQ，英国；地理与环境科学系，南安普顿大学，高菲尔德，南安普顿 SO17 1BJ，英国
Liqiu Meng，慕尼黑工业大学航天与测地系制图学教席，慕尼黑，德国
论文50
Boosting 3D point-based object detection by reducing information loss caused by discontinuous receptive fields
通过减少因不连续感受野造成的信息损失来提升3D点云目标检测
【摘要】
The point-based 3D object detection method is highly advantageous due to its lightweight nature and fast inference speed, making it a valuable asset in engineering fields such as intelligent transportation and autonomous driving. However, current advanced methods solely focus on learning features from the provided point cloud, neglecting the active role of unoccupied space. This results in the problem of discontinuous receptive field (DRF), leading to the loss of semantic and geometric information of the objects. To address this issue, we propose a new end-to-end single-stage point-based model, DRF-SSD, in this paper. DRF-SSD utilizes a PointNet++-style 3D backbone to maintain fast inference capability. Then, point-wise features are projected onto a plane in the Neck structure, and local and global information are aggregated through the designed Hierarchical Encoding–Decoding (HED) and Hybrid Transformer (HT) modules. The former fills in features for unoccupied space through convolutional layers, enhancing local features by interacting with features in occupied space during the learning process. The latter further expands the receptive field using the global learning ability of transformers. The spatial transformation and learning processes in HED and HT only involve key points, and HED is designed to have a special structure that maintains the sparsity of feature maps, preserving the efficiency of the model’s inference. Finally, query features are back-projected onto points for feature enhancement and input into the detection head for prediction. Extensive experiments on the KITTI datasets demonstrate that DRF-SSD achieves superior detection accuracy compared to previous methods, with significant improvements. Specifically, the approach obtains 2.25%, 0.66%, and 0.42% improvement for the metric of 3D Average Precision (AP3D
) under the easy, moderate, and hard settings, respectively. Additionally, the method enables other point-based detectors to achieve substantial gains, demonstrating its effectiveness. Our code will be made available at
https://github.com/AlanLiangC/DRF-SSD.git.
【摘要翻译】
基于点的三维物体检测方法因其轻量化特性和快速推理速度而具有很大优势，这使其在智能交通和自动驾驶等工程领域中成为宝贵的资产。然而，目前的先进方法仅专注于从提供的点云中学习特征，忽视了未占用空间的主动作用。这导致了不连续感受野（DRF）的问题，从而丧失了物体的语义和几何信息。为了解决这一问题，本文提出了一种新的端到端单阶段点云模型——DRF-SSD。DRF-SSD利用PointNet++风格的三维骨干网络来保持快速推理能力。然后，点特征被投影到颈部结构的平面上，通过设计的层次编码-解码（HED）和混合变换器（HT）模块聚合局部和全局信息。前者通过卷积层填充未占用空间的特征，通过与占用空间中的特征交互来增强局部特征。后者利用变换器的全局学习能力进一步扩展感受野。HED和HT中的空间变换和学习过程仅涉及关键点，HED被设计为具有特殊结构，以保持特征图的稀疏性，从而保持模型推理的效率。最后，查询特征被回投影到点上以增强特征，并输入到检测头中进行预测。对KITTI数据集的广泛实验表明，DRF-SSD在检测精度上显著优于以往的方法，具有显著改进。具体而言，该方法在简单、中等和困难设置下的三维平均精度（AP3D）指标分别提高了2.25%、0.66%和0.42%。此外，该方法使其他基于点的检测器也能取得显著增益，证明了其有效性。我们的代码将会发布在 https://github.com/AlanLiangC/DRF-SSD.git 上。
【doi】
https://doi.org/10.1016/j.jag.2024.104049
【作者信息】
Ao Liang, 中国科学院光电信息处理重点实验室，沈阳，110016，中国；中国科学院沈阳自动化研究所，沈阳，110016，中国；机器人与智能制造研究所，沈阳，110169，中国；中国科学院大学，北京，100049，中国
Haiyang Hua, 中国科学院光电信息处理重点实验室，沈阳，110016，中国；中国科学院沈阳自动化研究所，沈阳，110016，中国；机器人与智能制造研究所，沈阳，110169，中国；光学信息与仿真技术重点实验室，沈阳，110016，中国
Jian Fang, 中国科学院光电信息处理重点实验室，沈阳，110016，中国；中国科学院沈阳自动化研究所，沈阳，110016，中国；机器人与智能制造研究所，沈阳，110169，中国；光学信息与仿真技术重点实验室，沈阳，110016，中国
Huaici Zhao, 中国科学院光电信息处理重点实验室，沈阳，110016，中国；中国科学院沈阳自动化研究所，沈阳，110016，中国；机器人与智能制造研究所，沈阳，110169，中国；光学信息与仿真技术重点实验室，沈阳，110016，中国
Tianci Liu，中国科学院光电信息处理重点实验室，沈阳，110016，中国；中国科学院沈阳自动化研究所，沈阳，110016，中国；机器人与智能制造研究所，沈阳，110169，中国
论文51
A fully convolutional neural network model combined with a Hough transform to extract crop breeding field plots from UAV images
结合Hough变换的全卷积神经网络模型用于从无人机图像中提取作物育种田块
【摘要】
High-throughput phenotypic analysis plays an increasingly important role in crop breeding. In such research, the breeder usually establishes hundreds to thousands of plots, with each plot having its independent genetic breeding sources. The breeding plot extraction of genetic sources is usually outlined manually on RGB UAV imagery, which is time-consuming and subject to human bias. Therefore, a rapid method to extract the breeding plot for each genetic source in high-throughput phenotypic analysis would be very significant. In this paper, we propose a transferable method for extracting breeding plots from UAV RGB imagery. We utilized the fully convolutional neural network model A-UNet with an attention gate. After obtaining binary raster data from deep learning, we introduced post-processing. Subsequently, the raster data after post-processing were converted to vector data to obtain geographical coordinates. Finally, the UAV imagery was masked by the vector data to obtain the extraction results for each plot. The results showed that A-UNet achieved accuracies of over 90 % in precision, recall, and F1 score. Post-processing resulted in a 93 % average IoU in breeding plot extraction in the main study area. The average IOU achieved over 86 % in different spatial resolutions (1.6 cm and 0.4 cm), plot sizes (1 m× 1.5 and 2 m × 5 m), and crop types (rice). In summary, this study developed a method for extracting breeding plots in high-throughput phenotype analysis, which would help to be used as a high-throughput screening technique for accelerating crop breeding.
【摘要翻译】
高通量表型分析在作物育种中发挥着越来越重要的作用。在这类研究中，育种者通常会建立数百到数千个试验地块，每个地块都具有其独立的遗传育种来源。传统上，这些遗传来源的育种地块提取通常依赖于对RGB无人机影像的手工勾画，这不仅耗时而且容易受到人为偏差的影响。因此，开发一种快速提取每个遗传来源育种地块的方法在高通量表型分析中具有重要意义。本文提出了一种可转移的方法，用于从无人机RGB影像中提取育种地块。我们采用了带有注意力门的全卷积神经网络模型A-UNet。在获得深度学习生成的二值栅格数据后，我们进行了后处理。随后，将后处理后的栅格数据转换为矢量数据以获取地理坐标。最后，通过矢量数据对无人机影像进行掩膜，以获得每个地块的提取结果。结果表明，A-UNet在精度、召回率和F1分数上都达到了90%以上的准确率。后处理在主要研究区域内的育种地块提取中实现了93%的平均IoU。在不同空间分辨率（1.6厘米和0.4厘米）、地块大小（1米×1.5米和2米×5米）以及作物类型（稻米）下，平均IoU均超过86%。总之，本研究开发了一种用于高通量表型分析中提取育种地块的方法，可作为加速作物育种的高通量筛选技术。
【doi】
https://doi.org/10.1016/j.jag.2024.104057
【作者信息】
Xiaoxu Han, 国家信息农业工程技术中心 (NETCIA)，农业农村部作物系统分析与决策重点实验室，教育部智能农业工程研究中心，江苏省信息农业重点实验室，南京农业大学智能农业研究所，中国江苏省南京市卫岗1号，邮政编码210095
Meng Zhou, 国家信息农业工程技术中心（NETCIA）、农业农村部作物系统分析与决策重点实验室、教育部智能农业工程研究中心、江苏省信息农业重点实验室、南京农业大学智能农业研究所，中国江苏省南京市卫岗1号，邮政编码210095
Caili Guo, 国家信息农业工程技术中心（NETCIA）、农业农村部作物系统分析与决策重点实验室、教育部智能农业工程研究中心、江苏省信息农业重点实验室、南京农业大学智能农业研究所，中国江苏省南京市卫岗1号，邮政编码210095
Hongxu Ai, 国家信息农业工程技术中心（NETCIA）、农业农村部作物系统分析与决策重点实验室、教育部智能农业工程研究中心、江苏省信息农业重点实验室、南京农业大学智能农业研究所，中国江苏省南京市卫岗1号，邮政编码210095
Tongjie Li, 国家信息农业工程技术中心（NETCIA）、农业农村部作物系统分析与决策重点实验室、教育部智能农业工程研究中心、江苏省信息农业重点实验室、南京农业大学智能农业研究所，中国江苏省南京市卫岗1号，邮政编码210095
Wei Li, 国家工程技术中心信息农业（NETCIA），农业农村部作物系统分析与决策重点实验室（MARA），教育部智能农业工程研究中心（MOE），江苏省信息农业重点实验室，南京农业大学智能农业研究所，中国江苏省南京市卫岗1号，邮政编码210095
Xiaohu Zhang, 中国南京农业大学智能农业研究所，国家农业信息工程技术中心（NETCIA），农业农村部作物系统分析与决策关键实验室，教育部智能农业工程研究中心，江苏省信息农业重点实验室，江苏省南京市韦岗一号，邮政编码210095
Qi Chen, 美国夏威夷大学马诺阿分校地理与环境系，檀香山，夏威夷州，邮政编码96822
Chongya Jiang, 中国南京农业大学智能农业研究所，江苏省信息农业重点实验室，农业农村部作物系统分析与决策重点实验室，教育部智能农业工程研究中心，国家信息农业工程技术中心，南京市维岗一号，江苏省210095
Tao Cheng, 国家信息农业工程技术中心 (NETCIA)，农业农村部作物系统分析与决策重点实验室，教育部智能农业工程研究中心，江苏省信息农业重点实验室，智能农业研究所，南京农业大学，维岗一号，南京，江苏 210095，中国
Yan Zhu, 国家信息农业工程技术中心 (NETCIA)，农业农村部作物系统分析与决策重点实验室，教育部智能农业工程研究中心，江苏省信息农业重点实验室，智能农业研究所，南京农业大学，维岗一号，南京，江苏 210095，中国
Weixing Cao, 国家信息农业工程技术中心 (NETCIA)，农业农村部作物系统分析与决策重点实验室，教育部智能农业工程研究中心，江苏省信息农业重点实验室，智能农业研究所，南京农业大学，维岗一号，南京，江苏 210095，中国
Xia Yao，国家信息农业工程技术中心 (NETCIA)，农业农村部作物系统分析与决策重点实验室，教育部智能农业工程研究中心，江苏省信息农业重点实验室，智能农业研究所，南京农业大学，维岗一号，南京，江苏 210095，中国
论文52
Exploring the potential of multi-source satellite remote sensing in monitoring crop nutrient status: A multi-year case study of cranberries in Wisconsin, USA
探索多源卫星遥感在监测作物营养状态中的潜力：以美国威斯康星州蔓越莓为例的多年案例研究
【摘要】
A timing and precise diagnosis of crop nutrient status is essential for optimizing management practices that promote environmentally friendly and enhanced crop yields. Although plant tissue analysis has conventionally been employed to evaluate the nutritional status of crops, this method cannot capture the spatial variability of crop nutrients. In contrast, satellite-based remote sensing can monitor the nutrient status of crops across expansive areas. This study explored the capability of multi-source satellite images (PlanetScope-4: 3 m, 4 bands; PlanetScope-8: 3 m, 8 bands; Sentinel-2: 10–60 m, 13 bands; PRISMA: 30 m, 239 bands) in mapping 12 foliar nutrients in cranberries. Three machine learning approaches, including partial least squared regression (PLSR), support vector regression (SVR), random forest regression (RFR), were used to relate foliar nutrients to different types of satellite-derived features (SR: surface reflectance; VI: vegetation indices; TF: texture features) or their combinations (SR+VI, VI+TF and SR+VI+TF). Model performance was compared across different foliar nutrients, modelling approaches and combinations of model input features using R2 (the coefficient of determination) and RRMSE (relative root mean square error, = root mean square error/nutrient range × 100 %). Input features that were important to foliar nutrient modelling were identified. The model performance difference among nutrients was consistent between Planet-4 and Sentinel-2, as well as between Planet-8 and PRISMA. In the Planet-4 and Sentinel-2 derived models, N was best predicted (average R2 = 0.77, average RRMSE=15 %), followed by macronutrients S (0.60–0.63, 11 %), Mg (0.58–0.65, 10–11 %), Ca (0.49–0.51, 9 %), Na (0.69, 22 %), P (0.49, 9 %) and K (0.20, 8 %), and then by all micronutrients(i.e., Fe, Mn, B, Cu and Zn: R2 = 0.04–0.61; RRMSE=16–28 %). In the Planet-8 and PRISMA derived models, macronutrients (i.e., N, P, K, Mg, Ca, S and Na) had lower R2 and RRMSE (R2 = 0.06–0.59; RRMSE=7–57 %) than micronutrients (i.e., Fe, Mn, B, Cu and Zn: R2 = 0.18–0.60; RRMSE=19–66 %). The successful retrieval of foliar nutrients from satellite imagery was influenced by many factors, including the intercorrelation between nutrients and model input features, the data availability at critical growth stages, and satellite images characteristics (e.g., spatial and spectral resolutions). Except for foliar nitrogen, foliar nutrients typically do not exhibit distinct absorption features associated with C, H, N, or O molecular bonds in the 400–2500 nm range. Our results indicate that their successful retrieval can be primarily attributed to the association between foliar nutrients and other leaf components (e.g., pigments, water, and dry matter) that do display spectral features within this range. Our study demonstrated the potential of integrating multi-source satellite data for precise nutrient monitoring over large scales.
【摘要翻译】
作物养分状态的及时和准确诊断对于优化管理实践、促进环保和提高作物产量至关重要。尽管植物组织分析传统上用于评估作物的营养状态，但这种方法无法捕捉作物养分的空间变异性。相比之下，基于卫星的遥感可以监测广泛区域内的作物营养状态。本研究探讨了多源卫星影像（PlanetScope-4: 3 m, 4 波段；PlanetScope-8: 3 m, 8 波段；Sentinel-2: 10–60 m, 13 波段；PRISMA: 30 m, 239 波段）在绘制蔓越莓12种叶面养分方面的能力。使用了三种机器学习方法，包括偏最小二乘回归（PLSR）、支持向量回归（SVR）和随机森林回归（RFR），将叶面养分与不同类型的卫星衍生特征（SR: 表面反射率；VI: 植被指数；TF: 纹理特征）或它们的组合（SR+VI，VI+TF 和 SR+VI+TF）相关联。通过 R2（决定系数）和 RRMSE（相对均方根误差，即均方根误差/养分范围 × 100%）对不同叶面养分、建模方法和模型输入特征组合的模型性能进行了比较。识别了对叶面养分建模重要的输入特征。Planet-4 和 Sentinel-2 衍生模型中的养分模型性能差异一致，以及 Planet-8 和 PRISMA 之间的差异。在 Planet-4 和 Sentinel-2 衍生模型中，氮（N）的预测最佳（平均 R2 = 0.77，平均 RRMSE = 15%），其次是大养分硫（S）（0.60–0.63，11%）、镁（Mg）（0.58–0.65，10–11%）、钙（Ca）（0.49–0.51，9%）、钠（Na）（0.69，22%）、磷（P）（0.49，9%）和钾（K）（0.20，8%），然后是所有微量元素（即铁、锰、硼、铜和锌：R2 = 0.04–0.61；RRMSE = 16–28%）。在 Planet-8 和 PRISMA 衍生模型中，大养分（即氮、磷、钾、镁、钙、硫和钠）的 R2 和 RRMSE 较低（R2 = 0.06–0.59；RRMSE = 7–57%），而微量元素（即铁、锰、硼、铜和锌：R2 = 0.18–0.60；RRMSE = 19–66%）较高。卫星影像中叶面养分的成功检索受多种因素影响，包括养分和模型输入特征之间的相互关系、关键生长阶段的数据可用性以及卫星影像特征（例如空间和光谱分辨率）。除了叶面氮外，叶面养分在400–2500 nm范围内通常没有与 C、H、N 或 O 分子键相关的明显吸收特征。我们的结果表明，它们的成功检索主要归因于叶面养分与其他叶片成分（如色素、水分和干物质）之间的关联，这些成分在此范围内显示了光谱特征。我们的研究展示了整合多源卫星数据进行大尺度精确养分监测的潜力。
【doi】
https://doi.org/10.1016/j.jag.2024.104063
【作者信息】
Yurong Huang, 中国，广东省广州市510275，中山大学地理与规划学院，广东省城市化与地理模拟省重点实验室，北方喀斯特地区碳水研究站
Nanfeng Liu, 中国，510275，广州市，中山大学地理与规划学院，广东省城市化与地理模拟省重点实验室，北方喀斯特地区碳水研究站
Erin Wagner Hokanson, 美国，威斯康星州麦迪逊市，53715，威斯康星大学麦迪逊分校，空间科学与工程中心
Nicole Hansen, 美国，威斯康星州内塞达市，54646，克兰贝里克溪蔓越莓公司
Philip A. Townsend，美国威斯康星大学麦迪逊分校，森林与野生生物生态系，威斯康星州麦迪逊市，邮政编码53705
论文53
Insight into large-scale LULC changes and their drivers through breakpoint characterization – An application to Senegal
通过断点特征化洞察大规模土地利用/土地覆盖变化及其驱动因素——应用于塞内加尔
【摘要】
As global land cover/ land use change (LULCC) threatens the human’s well-being, accurate detection and characterization of LULCC is of paramount importance. The increasing availability of dense satellite image time series (SITS), together with the ever-improving change detection algorithms, has allowed significant progress to be made. However, much remains to be done in its characterization.This study aims to uncover potential relationships between changes in Normalized Difference Vegetation Index (NDVI) SITS patterns and their drivers. It distinguishes itself by representing phenological changes not only as transitions between specific patterns, but also by examining the nature of these changes—whether abrupt, gradual, or seasonal. For seasonal changes, it further refines the analysis to determine their impact on the amplitude, number of seasons (NOS), or length of seasons (LOS) components. Our focus is to provide insights into the land dynamics and drivers of change in Senegal using an RGB (red, green, blue) composite change map. This map is derived from three MODIS NDVI time series change metrics detected by BFASTm-L2 within the MODIS NDVI 2000–2021 SITS: magnitude of change, direction of change, and dissimilarity of time series shape. The 250-meter resolution MODIS data served as an optimal data source for this analysis due to its high temporal resolution (near daily) and extensive coverage over 20 years.The sensitivity of each metric to different types of change was first tested on a simulated dataset before being applied to the MODIS SITS. The RGB change map enabled visualization of different “signatures” of change, which, combined with ground information, rainfall data, NDVI time series analysis, and Google Earth imagery, helped link these signatures to various drivers of change. Climatic and anthropogenic changes, such as those induced by Large Scale Agricultural Investments (LSAI) or mining, were visually inferred from the RGB map.This approach demonstrates the usefulness of integrating the type of change, especially seasonal change, into the characterization of land change. This method has the advantage of being fast, interpretable, robust to noise and easily transferable to different regions.
【摘要翻译】
全球土地覆盖/土地使用变化（LULCC）威胁着人类的福祉，因此准确检测和描述LULCC至关重要。随着密集卫星图像时间序列（SITS）的日益普及以及变化检测算法的不断改进，取得了显著进展。然而，在对这些变化的特征化方面，仍有许多工作要做。本研究旨在揭示归一化植被指数（NDVI）时间序列模式变化与其驱动因素之间的潜在关系。研究的特点在于不仅将物候变化表示为特定模式之间的过渡，还考察这些变化的性质——是突然的、渐进的还是季节性的。对于季节性变化，进一步细化分析以确定其对幅度、季节数量（NOS）或季节长度（LOS）分量的影响。我们的重点是使用RGB（红色、绿色、蓝色）复合变化图提供关于塞内加尔土地动态和变化驱动因素的见解。该图是通过BFASTm-L2在MODIS NDVI 2000–2021时间序列中检测到的三个变化度量（变化幅度、变化方向和时间序列形状的差异）得出的。250米分辨率的MODIS数据由于其高时间分辨率（接近每日）和覆盖范围广泛（超过20年）而成为此次分析的理想数据源。首先，在模拟数据集上测试了每个度量对不同类型变化的敏感性，然后将其应用于MODIS时间序列。RGB变化图使不同“变化特征”的可视化成为可能，结合地面信息、降水数据、NDVI时间序列分析和Google Earth影像，帮助将这些特征与各种变化驱动因素联系起来。气候和人为变化（例如，大规模农业投资（LSAI）或采矿）在RGB图中被直观推断出来。这种方法展示了将变化类型，特别是季节性变化，整合到土地变化特征化中的有用性。这种方法的优点是快速、易于解释、对噪声具有鲁棒性，并且容易转移到不同区域。
【doi】
https://doi.org/10.1016/j.jag.2024.104066
【作者信息】
Yasmine Ngadi Scarpetta, 法国，蒙彼利埃，34093，尚-弗朗索瓦·布雷顿街500号，遥感大楼，Cirad，UMR TETIS；法国，蒙彼利埃，F-34398，UMR TETIS AgroParisTech CIRAD CNRS INRAE 蒙彼利埃大学；法国，蒙彼利埃，F-34398，UMR ESPACE-DEV，IRD，蒙彼利埃大学，安提利大学，法属圭亚那大学，留尼汪大学
Valentine Lebourgeois, 法国，蒙彼利埃，34093，尚-弗朗索瓦·布雷顿街500号，遥感大楼，Cirad，UMR TETIS；法国，蒙彼利埃，F-34398，UMR TETIS，AgroParisTech，CIRAD，CNRS，INRAE，蒙彼利埃大学
Mohamadou Dieye, ISRA-BAME，石油路线，BP 3120 达喀尔，塞内加尔
Anne-Elisabeth Laques, IRD UMR ESPACE-DEV，IRD，蒙彼利埃大学，安的列斯大学，法属圭亚那大学，留尼汪大学，塔那那利佛，马达加斯加
Agnès Begue，Cirad，UMR TETIS，遥感大楼，让-弗朗索瓦·布雷东街500号，34093 蒙彼利埃，法国；UMR TETIS AgroParisTech CIRAD CNRS INRAE 蒙彼利埃大学，F-34398 蒙彼利埃，法国
论文54
Accelerated forest modeling from tree canopy point clouds via deep learning
通过深度学习加速从树冠点云进行森林建模
【摘要】
Rapid generation of tree models from point clouds of tree canopies holds wide-ranging applications in the field of earth sciences, including forest ecology research, environmental monitoring, and forest management. Traditional tree modeling methods rely on procedural models to simulate tree growth, which are time-consuming due to their extensive manual parameterization. Furthermore, existing deep learning methods struggle to generate visually realistic tree models because of the complex branch structures and specific natural patterns of trees. To address these challenges, this paper proposes a novel deep learning-based method for rapidly generating tree models that align with the shape of the tree canopy. Different from traditional methods, we use deep neural networks to build branch graphs for generating tree models. Our method consists of two main steps: i) the 3D coordinates of each tree node are generated from the canopy point cloud by the designed node coordinate generation network; ii) a graph neural network is proposed to predict node attributes and the adjacency relationship between nodes. To form the tree structure, the discrete nodes are connected by using the minimum spanning tree algorithm combined with the adjacency relationship. The attributes of the node include width, whether it is a leaf node, and leaf node size, which are used for subsequent construction of the tree’s mesh. To validate the effectiveness of our proposed method, a large-scale dataset containing 10 forests with 3216 tree canopies is constructed and open sourced for the study of generating tree models from point clouds of tree canopies. Experimental results demonstrate our method’s efficiency in generating tree models quickly (reducing the average canopy-to-tree reconstruction time from 7 min to less than 0.5 s) while preserving visual authenticity and accurately matching tree canopy shapes, making it suitable for a wide range of forest reconstructions.
【摘要翻译】
快速生成树冠点云的树木模型在地球科学领域具有广泛的应用，包括森林生态研究、环境监测和森林管理。传统的树木建模方法依赖于程序化模型来模拟树木生长，这些方法由于需要大量手动参数设置而耗时。此外，现有的深度学习方法在生成视觉上逼真的树木模型时面临困难，因为树木的分支结构复杂且具有特定的自然模式。为了解决这些挑战，本文提出了一种新颖的基于深度学习的方法，用于快速生成与树冠形状一致的树木模型。与传统方法不同，我们使用深度神经网络来构建分支图以生成树木模型。我们的方法包括两个主要步骤：i) 通过设计的节点坐标生成网络从树冠点云中生成每个树节点的三维坐标；ii) 提出一种图神经网络来预测节点属性以及节点之间的邻接关系。为了形成树结构，使用最小生成树算法结合邻接关系连接离散节点。节点属性包括宽度、是否是叶节点以及叶节点大小，这些属性用于后续构建树的网格。为了验证我们提出方法的有效性，我们构建了一个包含10个森林、3216个树冠的大规模数据集，并开放用于从树冠点云生成树木模型的研究。实验结果表明，我们的方法在快速生成树木模型方面效率高（将平均树冠重建时间从7分钟减少到不到0.5秒），同时保持视觉真实性并准确匹配树冠形状，使其适用于广泛的森林重建任务。
【doi】
https://doi.org/10.1016/j.jag.2024.104074
【作者信息】
Jiabo Xu, 遥感与信息工程学院，武汉大学，武汉，湖北，430079，中国
Zhili Zhang, 遥感与信息工程学院，武汉大学，武汉，湖北，430079，中国
Xiangyun Hu, 遥感与信息工程学院，武汉大学，武汉，湖北，430079，中国；湖北罗嘉实验室，武汉大学，武汉，湖北，430079，中国
Tao Ke，遥感与信息工程学院，武汉大学，武汉，湖北，430079，中国；湖北罗嘉实验室，武汉大学，武汉，湖北，430079，中国
论文55
Automatic mapping of aquaculture activity in the Atlantic Ocean
自动化绘制大西洋地区的水产养殖活动
【摘要】
The production of wild fish has remained relatively stable in the last two decades, whereas aquaculture organism production has increased to the point where it has exceeded wild catches. In this context, accurate and up-to-date information about the current usage of marine areas for aquaculture is crucial for the planning of marine activities. However, this data is often limited to national authorities, and discrepancies between planned and real practices can arise in available data. In this study, a novel methodology to automatically map and verify the current activity of aquaculture crops across European regions based on freely available satellite data is proposed. The European Space Agency’s (ESA) Sentinel-1 mission provides Synthetic Aperture Radar (SAR) images, which serve as the basis for the analysis. Multiple SAR images of the same locations are processed using ESA Sentinel Application Platform (SNAP) software and merged to remove temporal noise-like artifacts caused by factors such as ships and waves. Next, the iDPolRAD algorithm is employed to detect potential aquaculture sites, which initially include noise from coastal zones and unwanted human and natural structures that pass through the filter. The aquaculture sites are classified using a ResNet18 model with 93% of the sites correctly classified. This implies that it is feasible to monitor marine areas using satellite radar data to track aquaculture areas. However, generalization power across regions is poor likely due to the diversity of types of structures used and species cultivated. Further studies are needed to investigate factors influencing the detectability of different aquaculture sites such as cage geometry or SAR image resolution in order to enhance the accuracy and comprehensiveness of the mapping process. This study highlights the potential of SAR data, coupled with image processing and classification techniques, as a viable means to map large marine areas dedicated to aquaculture.
【摘要翻译】
野生鱼类的产量在过去二十年中保持相对稳定，而水产养殖生物的产量已增加到超过了野生捕捞。在这种背景下，准确且最新的海洋养殖区使用信息对海洋活动的规划至关重要。然而，这些数据通常仅限于国家当局，并且计划与实际做法之间可能会存在差异。本研究提出了一种基于自由获取的卫星数据自动绘制和验证欧洲地区水产养殖作物当前活动的新方法。欧洲空间局（ESA）的Sentinel-1任务提供了合成孔径雷达（SAR）图像，作为分析的基础。对相同位置的多个SAR图像进行处理，使用ESA Sentinel应用平台（SNAP）软件进行合并，以去除由船只和波浪等因素引起的时间噪声样伪影。接下来，使用iDPolRAD算法检测潜在的水产养殖场所，这些场所最初包括来自沿海区域的噪声以及通过筛选的未被需要的人为和自然结构。使用ResNet18模型对水产养殖场所进行分类，其中93%的场所被正确分类。这表明使用卫星雷达数据监测海洋区域以跟踪水产养殖区是可行的。然而，由于所使用结构类型和养殖物种的多样性，区域间的泛化能力较差。需要进一步研究影响不同水产养殖场所可检测性的因素，例如笼体几何形状或SAR图像分辨率，以提高绘图过程的准确性和全面性。本研究突出了SAR数据与图像处理和分类技术结合作为绘制大型海洋水产养殖区的可行方法的潜力。
【doi】
https://doi.org/10.1016/j.jag.2024.104061
【作者信息】
Xabier Lekunberri, AZTI，海洋研究，巴斯克研究与技术联盟（BRTA），Txatxarramendi Ugartea z/g，Sukarrieta，Bizkaia 48395，西班牙；巴斯克国家大学（UPV/EHU），圣塞巴斯蒂安，西班牙
J. David Ballester-Berman, 计算机研究所（IUII），阿利坎特大学，西班牙
Ignacio Arganda-Carreras, 巴斯克大学（UPV/EHU），圣塞巴斯蒂安，西班牙；IKERBASQUE，巴斯克科学基金会，比尔巴鄂，西班牙；圣塞巴斯蒂安国际物理中心（DIPC），圣塞巴斯蒂安，西班牙；生物物理学研究所（CSIC，UPV/EHU），莱奥伊亚，西班牙
Jose A. Fernandes-Salvador，AZTI, 海洋研究，巴斯克研究与技术联盟（BRTA），Txatxarramendi Ugartea z/g, Sukarrieta, Bizkaia 48395, 西班牙
论文56
Incorporating fire spread simulation and machine learning algorithms to estimate crown fire potential for pine forests in Sichuan, China
结合火灾扩散模拟和机器学习算法估计四川省松林的冠层火灾潜力
【摘要】
Accurate estimation of crown fire potential (CFP) can improve guidance on crown fire control and management. However, robust simulations of crown fire behavior are still challenging, limiting the accuracy of regional-scale CFP mapping. This study aims to incorporate fire spread simulation and machine learning algorithms to improve CFP mapping at a regional scale. First, we built a crown fire dataset using the fire simulations from the FARSITE model, as well as multi-source data, including fuel, weather, and topography variables. Fuel model parameters were optimized with four metaheuristic algorithms for robust fire simulations. Then, the hybrid models of CFP estimation (TBA-ML) were established by coupling with the transfer AdaBoost (TrAdaBoost) algorithm and three machine learning (ML) algorithms, i.e., Bayesian Network (BN), Random Forest (RF), and Support Vector Machine (SVM), to estimate CFP for crown fire danger assessment spatially. Results showed that the TBA-BN model performed best in estimating CFP with higher accuracy (AUC>0.9 and F1 score > 0.8) than the RF- and SVM-based CFP models. The variable importance and causal analysis showed that fuel and topography variables have major contributions to crown fire occurrence. Finally, we mapped monthly average passive and active CFP at regional scales and qualitatively demonstrated that our CFP time-series products successfully captured the dynamic change of crown fire danger. The above results suggest the potential of integrating fire spread simulation and machine learning algorithms to accurately estimate CFP to improve crown fire management.
【摘要翻译】
准确估计冠层火灾潜力（CFP）可以改善对冠层火灾的控制和管理。然而，冠层火灾行为的可靠模拟仍然具有挑战性，限制了区域尺度CFP绘制的准确性。本研究旨在结合火灾扩散模拟和机器学习算法，以提高区域尺度的CFP绘制。首先，我们使用FARSITE模型的火灾模拟数据和包括燃料、天气和地形变量在内的多源数据构建了一个冠层火灾数据集。通过四种元启发式算法优化了燃料模型参数，以实现稳健的火灾模拟。然后，结合转移AdaBoost（TrAdaBoost）算法和三种机器学习（ML）算法，即贝叶斯网络（BN）、随机森林（RF）和支持向量机（SVM），建立了CFP估计的混合模型（TBA-ML），以空间方式估计CFP以评估冠层火灾危险性。结果表明，TBA-BN模型在CFP估计中表现最佳，具有比基于RF和SVM的CFP模型更高的准确性（AUC>0.9和F1分数>0.8）。变量重要性和因果分析表明，燃料和地形变量对冠层火灾的发生有主要贡献。最后，我们绘制了区域尺度上每月平均的被动和主动CFP，并定性展示了我们的CFP时间序列产品成功捕捉了冠层火灾危险的动态变化。上述结果表明，结合火灾扩散模拟和机器学习算法在准确估计CFP以改善冠层火灾管理方面具有潜力。
【doi】
https://doi.org/10.1016/j.jag.2024.104080
【作者信息】
Rui Chen, 电子科技大学资源与环境学院，成都 611731，中国
Binbin He, 电子科技大学资源与环境学院，成都 611731，中国
Yanxi Li, 电子科技大学资源与环境学院，成都 611731，中国
Yiru Zhang, 电子科技大学资源与环境学院，成都 611731，中国
Zhanmang Liao, 电子科技大学资源与环境学院，成都 611731，中国
Chunquan Fan, 电子科技大学资源与环境学院，成都 611731，中国
Jianpeng Yin, 电子科技大学资源与环境学院，成都 611731，中国
Hongguo Zhang，电子科技大学资源与环境学院，成都 611731，中国
论文57
A hybrid model coupling PROSAIL and continuous wavelet transform based on multi-angle hyperspectral data improves maize chlorophyll retrieval
基于多角度高光谱数据的PROSAIL与连续小波变换耦合的混合模型改进玉米叶绿素提取
【摘要】
Chlorophyll is both a cornerstone of plant photosynthesis and an important indicator for assessing crop growth and health. Although many previous studies have explored the use of remote sensing to retrieve chlorophyll content, there is room for improvement in the proposed retrieval models, especially the hybrid model, and its performance in combination with multi-angle remote sensing remains unknown. To this end, we developed a hybrid chlorophyll retrieval model by coupling PROSAIL, Gaussian process regression, and continuous wavelet transform (CWT) based on multi-angle (−60° to 60°) hyperspectral observations of maize. The CWT converts PROSAIL-modeled and measured spectral reflectance into wavelet features (WF) that finely capture signals due to chlorophyll changes, making WF-based hybrid models (HMWF) promising for enhanced chlorophyll retrieval. Our results show that for leaf chlorophyll content (LCC) and canopy chlorophyll content (CCC) retrieval, combining low and medium scale WFs (scales3-5) with hybrid models is more advantageous than using other scale WFs. The accuracy of the HMWF based on the best-scale WF was significantly higher than that of the hybrid model based on original spectrum or vegetation indices. Additionally, our evaluation of the effect of viewing zenith angles (VZAs) on HMWF showed that the accuracies of HMWF acquired at non-nadir angles were generally higher than those acquired at nadir angle. Among all models, the HMWF based on the scale3 WF had the highest accuracy at −10°, with R2 = 0.85 and RMSE=3.55 for LCC retrievals, and R2 = 0.78 and RMSE=0.22 for CCC retrievals. Furthermore, the HMWF showed the least sensitivity to changes in VZAs, especially in the range of −10° to −40°. Overall, these findings highlight the effectiveness of HMWF with multi-angle hyperspectral data in improving chlorophyll retrieval accuracy. This study serves as a reference for crop parameter retrieval, crucial for advancing agricultural monitoring and management.
【摘要翻译】
叶绿素是植物光合作用的基石，也是评估作物生长和健康的重要指标。尽管许多先前的研究探讨了使用遥感技术获取叶绿素含量，但现有的检索模型，特别是混合模型的表现还有改进的空间，尤其是在与多角度遥感结合时的表现尚不清楚。为此，我们开发了一种混合叶绿素检索模型，该模型结合了PROSAIL模型、高斯过程回归和基于多角度（−60°至60°）高光谱观测的连续小波变换（CWT）。CWT将PROSAIL模拟和测量的光谱反射率转换为小波特征（WF），这些特征精确地捕捉了由叶绿素变化引起的信号，使基于WF的混合模型（HMWF）在增强叶绿素检索方面具有很大的潜力。我们的结果显示，对于叶片叶绿素含量（LCC）和冠层叶绿素含量（CCC）检索，将低和中尺度WF（尺度3-5）与混合模型结合使用，比使用其他尺度WF更具优势。基于最佳尺度WF的HMWF的准确性显著高于基于原始光谱或植被指数的混合模型。此外，我们评估了视天顶角（VZA）对HMWF的影响，结果显示，非天顶角获得的HMWF准确性通常高于天顶角获得的准确性。在所有模型中，基于尺度3 WF的HMWF在−10°角度下表现出最高的准确性，LCC检索的R2为0.85，RMSE为3.55；CCC检索的R2为0.78，RMSE为0.22。此外，HMWF对VZA变化的敏感性最低，尤其是在−10°至−40°范围内。总体而言，这些发现突显了基于多角度高光谱数据的HMWF在提高叶绿素检索准确性方面的有效性。本研究为作物参数检索提供了参考，这对于推进农业监测和管理至关重要。
【doi】
https://doi.org/10.1016/j.jag.2024.104076
【作者信息】
Anting Guo, 中国科学院航空航天信息研究所遥感与数字地球重点实验室，北京 100094，中国
Wenjiang Huang, 中国科学院航空航天信息研究所遥感与数字地球重点实验室，北京 100094，中国；中国科学院大学，北京 100049，中国
Binxiang Qian, 中国科学院航空航天信息研究所遥感与数字地球重点实验室，北京 100094，中国；中国科学院大学，北京 100049，中国
Huichun Ye, 中国科学院航空航天信息研究所数字地球科学重点实验室，北京 100094，中国
Quanjun Jiao, 中国科学院航空航天信息研究所遥感与数字地球重点实验室，北京 100094，中国
Xiangzhe Cheng, 中国科学院航空航天信息研究所遥感与数字地球重点实验室，北京 100094，中国；中国科学院大学，北京 100049，中国
Chao Ruan，中国合肥 230601，安徽大学互联网学院
论文58
Large kernel convolution application for land cover change detection of remote sensing images
大核卷积应用于遥感图像的土地覆盖变化检测
【摘要】
In land cover change detection tasks, extracting universal features of changing targets is crucial for achieving precise detection results. A larger receptive field helps the model capture these universal features of changing targets. Although Large Kernel Convolution has been widely used in the field of computer vision, its potential in land cover change detection of remote sensing images has not been fully explored. To address this, a novel Re-parameterization Large kernel Convolution Network for Change Detection (CD-RLKNet) is proposed. CD-RLKNet utilizes Spatial and Temporal Adaptive Fusion Module to preliminarily extract spatial–temporal features from bi-temporal remote sensing images, resulting in a coarse-grained fused feature map. Features Assimilation Assistant Module extracts independent features from land cover information of each temporal, serving as auxiliary information for fine-grained features. Bi-temporal Features Integration Module utilizes large kernel convolution to extract bi-temporal land cover features with a larger receptive field, capturing fine-grained differences in these features. Experiments have been conducted on SYSU-CD, LEVIR-CD and GVLM-CD datasets, and results show that the proposed CD-RLKNet achieves IoU values of 0.6882, 0.8294 and 0.7729, respectively, surpassing the compared SOTA models. The code of CD-RLKNet can be achieved from https://github.com/juncyan/cdrlknet.git.
【摘要翻译】
在土地覆盖变化检测任务中，提取变化目标的普遍特征对于实现精确检测结果至关重要。较大的感受野有助于模型捕捉这些变化目标的普遍特征。虽然大核卷积（Large Kernel Convolution）在计算机视觉领域已被广泛使用，但其在遥感图像的土地覆盖变化检测中的潜力尚未得到充分探索。为了解决这一问题，本文提出了一种新型的重参数化大核卷积网络（CD-RLKNet）用于变化检测。CD-RLKNet利用空间和时间自适应融合模块（Spatial and Temporal Adaptive Fusion Module）初步提取双时相遥感图像的空间–时间特征，生成粗粒度的融合特征图。特征融合助手模块（Features Assimilation Assistant Module）从每个时相的土地覆盖信息中提取独立特征，作为细粒度特征的辅助信息。双时相特征集成模块（Bi-temporal Features Integration Module）利用大核卷积提取双时相土地覆盖特征，具有更大的感受野，捕捉这些特征中的细粒度差异。实验在SYSU-CD、LEVIR-CD和GVLM-CD数据集上进行，结果表明，所提出的CD-RLKNet在这些数据集上的IoU值分别为0.6882、0.8294和0.7729，优于对比的SOTA模型。CD-RLKNet的代码可以从(https://github.com/juncyan/cdrlknet.git)获取。
【doi】
https://doi.org/10.1016/j.jag.2024.104077
【作者信息】
Junqing Huang, 中国澳门特别行政区999078，澳门理工大学应用科学学院
Xiaochen Yuan, 应用科学学院，澳门理工大学，中国澳门特别行政区 999078
Chan-Tong Lam, 应用科学学院，澳门理工大学，中国澳门特别行政区 999078
Wei Ke, 应用科学学院，澳门理工大学，中国澳门特别行政区 999078
Guoheng Huang，计算机科学与技术学院，广东工业大学，中国广州 510000
论文60
Retrieving heavy metal concentrations in urban soil using satellite hyperspectral imagery
使用卫星高光谱影像检索城市土壤中重金属浓度
【摘要】
Efficient prediction and precise depiction of heavy metal concentrations in urban soil are essential for mitigating non-point source pollution and safeguarding public health. Therefore, this research investigated the estimation of soil heavy metal concentrations derived from Gaofen-5 (GF-5) hyperspectral images calibrated by the direct standardization (DS) algorithm. The inversion strategy for soil heavy metal concentrations in response to the two-dimensional soil spectral index (2D-SSI) was proposed by coupling Pearson correlation coefficient (r) and competitive adaptive reweighting algorithm (CARS) for feature selection. The results indicated that the optimal models based on 2D-SSI outperform the models based on calibrated, filtered original spectral bands. For Pb, Cu, Cd, and Hg, the optimal model determination coefficients for the validation data set (R2v) were 0.871 (SVM), 0.883 (BPNN), 0.834 (PLSR), and 0.907 (PLSR), respectively. The spectral features were highlighted in the two-dimensional feature space, and the predicted distribution of heavy metal concentrations was aligned with the observed ground measurements. This study revealed that the prediction strategy based on DS-corrected GF-5 AHSI images with constructed 2D-SSI features can serve as a reliable technical approach for soil heavy metal prediction and pollution prevention.
【摘要翻译】
高效预测和精确描绘城市土壤中的重金属浓度对于减轻非点源污染和保护公共健康至关重要。因此，本研究探讨了基于高分五号（GF-5）高光谱图像和直接标准化（DS）算法校准的土壤重金属浓度估计。提出了一种基于二维土壤光谱指数（2D-SSI）的土壤重金属浓度反演策略，通过耦合皮尔逊相关系数（r）和竞争自适应重加权算法（CARS）进行特征选择。结果表明，基于2D-SSI的最佳模型优于基于校准和滤波后的原始光谱波段的模型。对于Pb、Cu、Cd和Hg，验证数据集的最佳模型决定系数（R²v）分别为0.871（SVM）、0.883（BPNN）、0.834（PLSR）和0.907（PLSR）。在二维特征空间中突出了光谱特征，预测的重金属浓度分布与实际地面测量结果一致。该研究揭示了基于DS校正的GF-5 AHSI图像和构建的2D-SSI特征的预测策略可以作为土壤重金属预测和污染防治的可靠技术方法。
【doi】
https://doi.org/10.1016/j.jag.2024.104079
【作者信息】
Nannan Yang, 西安长安大学土地工程学院，中国陕西省西安市710064
Liangzhi Li, 西安长安大学土地工程学院，中国陕西省西安市710064；西安长安大学土地工程学院，西安城市空间信息重点实验室，中国陕西省西安市710064；西安长安大学陕西省土地复垦工程重点实验室，中国陕西省西安市710064
Ling Han, 长安大学土地工程学院，中国陕西省西安市710064；长安大学土地工程学院西安城市空间信息重点实验室，中国陕西省西安市710064；长安大学陕西省土地复垦工程重点实验室，中国陕西省西安市710064
Kyle Gao, 加拿大滑铁卢大学系统设计工程系，滑铁卢，ON N2L 3G1
Songjie Qu, 中国，陕西省，西安，710064，长安大学，土地工程学院
Jonathan Li，加拿大，安大略省，滑铁卢，N2L 3G1，滑铁卢大学，系统设计工程系；加拿大，安大略省，滑铁卢，N2L 3G1，滑铁卢大学，地理与环境管理系
论文61
Cross-temporal and spatial information fusion for multi-task building change detection using multi-temporal optical imagery
跨时间和空间信息融合用于多任务建筑变化检测：基于多时相光学影像的应用
【摘要】
Accurate detection of changes in buildings is crucial for the understanding of urban development. The growing accessibility of remote sensing imagery has enabled urban scale change detection (CD) in both 2D and 3D. However, existing methods have not yet fully exploited the fusion of feature information in multi-temporal images, resulting in insufficient accuracy in 2D changed regions or in elevation changes. To this end, a Cross-temporal and Spatial Context Learning Network (CSCLNet) aimed at multi-task building CD from dual-temporal optical images is proposed, capturing both 2D and 3D changes simultaneously. It leverages a CNN network to extract multi-layer semantic features. Subsequently, two modules, Cross-temporal Transformer Semantic Enhancement (CTSE) and Multi-layer Feature Fusion (MFF), are developed to refine the feature representations. CTSE enhances temporal information by cross attention of dual-temporal features to enable interactions and MFF fuses multi-layer features and enhances attention to global and local spatial context. Finally, two prediction heads are introduced to separately handle 2D and 3D change prediction, identifying changed building objects and their elevation changes. Experiments conducted with two public datasets, 3DCD and SMARS, show that the CSCLNet achieves state-of-the-art for both 2D and 3D CD tasks. In particular, the change-specific RMSE of elevation changes has been reduced to 4.52 m in real world scenes. The code is available at: https://github.com/Geo3DSmart/CSCLNet.
【摘要翻译】
准确检测建筑物变化对于理解城市发展至关重要。遥感影像的日益普及使得城市尺度的变化检测（CD）在2D和3D领域成为可能。然而，现有方法尚未充分利用多时相图像中的特征信息融合，导致2D变化区域或高度变化的准确性不足。为此，提出了一种跨时相和空间上下文学习网络（CSCLNet），旨在通过双时相光学图像实现多任务建筑变化检测，同时捕捉2D和3D变化。该方法利用CNN网络提取多层语义特征。随后，开发了两个模块：跨时相变换器语义增强（CTSE）和多层特征融合（MFF），以细化特征表示。CTSE通过双时相特征的交叉注意力增强时序信息，从而实现互动；MFF则融合多层特征，并增强对全局和局部空间上下文的关注。最后，引入两个预测头分别处理2D和3D变化预测，识别变化的建筑物对象及其高度变化。通过在两个公开数据集3DCD和SMARS上的实验，CSCLNet在2D和3D变化检测任务中均达到了最新的研究水平。特别是在实际场景中，高度变化的特定RMSE已降至4.52米。代码可在以下链接获取：https://github.com/Geo3DSmart/CSCLNet。
【doi】
https://doi.org/10.1016/j.jag.2024.104075
【作者信息】
Wen Xiao, 中国地质大学地理与信息工程学院，430074 武汉，中国；中国地质大学国家地理信息系统工程技术研究中心，430074武汉，中国
Hui Cao, 中国地质大学地理与信息工程学院，430074 武汉，中国
Yuqi Lei, 中国地质大学（武汉）地理与信息工程学院，430074 武汉，中国
Qiqi Zhu, 地理与信息工程学院，中国地质大学（武汉），430074 武汉，中国；国家地理信息系统工程技术研究中心，中国地质大学（武汉），430074 武汉，中国
Nengcheng Chen，国家地理信息系统工程技术研究中心，中国地质大学（武汉），430074 武汉，中国
论文62
Estimating terrain elevations at 10 m resolution by Integrating random forest machine learning model and ICESat-2, Sentinel-1, and Sentinel-2 satellite remotely sensed data
通过整合随机森林机器学习模型与ICESat-2、Sentinel-1和Sentinel-2卫星遥感数据以10米分辨率估算地形高程
【摘要】
Accurate mapping of terrain elevations at a large scale and fine resolution can characterize the detailed surface height and geomorphic changes and is very critical for the studies of the internal motions and external forces of the earth. The emergence of the Ice, Cloud, and Land Elevation Satellite-2 (ICESat-2) offers unprecedented possibilities for global elevation mapping with high vertical accuracy using three-dimensional photon points. However, the ICESat-2 photon points are still sparse in terms of spatial/horizontal resolution, making it unable to satisfy the high-resolution demand of terrain elevation mapping and digital elevation model production. A few previous studies have attempted to estimate elevations/topography in regions with single landscape and landcover (e.g., forest, shallow water, and polar regions) by combining ICESat-2 data with other passive satellite remotely sensed data. However, the potential and capability of ICESat-2 for mapping elevations for spatially continuous large regions with multiple complicated land cover types remains unknown. In this study, a spatially continuous large-scale terrain elevation estimation method is developed under multiple land covers based on the random forest model and the freely accessed satellite data of ICESat-2, Sentinel-1, and Sentinel-2. The core principle is to construct a random forest model that can characterize the complicated relationships of the ICESat-2 ATL03 terrain elevations and their corresponding land cover related polarization characteristics and spectral variables from Sentinel-1 and Sentinel-2, respectively. Integrating the superiorities of the data of these three different satellites enables the proposed method to extrapolate the terrain elevations with decimeter-level vertical accuracy and 10 m spatial/horizontal resolution simultaneously without any prior in situ data or manually set parameters. The proposed method is tested using the elevations from 2021 to 2022 at the third largest island (Chongming Island, Shanghai) in China. The estimated terrain elevations are locally validated with the airborne LiDAR-derived elevations. Moreover, they are compared with the ICESat-2 ATL08 height_terrain_bestfit data and Global Ecosystem Dynamics Investigation L2A elev_lowestmode data from the global perspectives. The predicted elevations exhibit a high correlation with the measured elevations from the two airborne LiDAR validation regions with root mean square errors (RMSE) of 0.34 and 0.59 m. The averaged RMSEs of the predicted elevations at different land covers are 1.26 and 1.18 m when compared with those derived from ATL08 and GEDI L2A, respectively. No remarkable abnormal predicted elevations are observed. This finding suggests the satisfactory robustness performance of the proposed method under different land covers and a relatively good consistency between the predicted elevations and the actual terrain of the entire island. As far as we know, the present work is the first to map elevations at 10 m resolution based only on the newly available satellite active and passive remotely sensed data without any ground truth surveys, manual intervention, and prior knowledge. Different with existing studies for terrain elevation mapping only at single landcovers, the proposed method demonstrates the capability and effectiveness of ICESat-2 for any landforms and landcovers and shows great potential for high-accuracy and high-resolution time-series terrain elevation estimation and updating at regional/national/global scales.
【摘要翻译】
大规模高分辨率地形高程的精确测绘对于了解地表高度和地貌变化至关重要，并且对于研究地球内部运动和外部力量具有重要意义。冰云陆地高程卫星-2（ICESat-2）的出现，为全球高精度垂直高程测绘提供了前所未有的可能性，该卫星利用三维光子点进行高精度测量。然而，ICESat-2的光子点在空间/水平分辨率方面仍然稀疏，无法满足高分辨率地形高程测绘和数字高程模型制作的需求。一些先前的研究尝试通过结合ICESat-2数据与其他被动卫星遥感数据，在单一景观和地貌（如森林、浅水区和极地地区）中估算高程/地形。然而，ICESat-2在多个复杂地貌类型的大面积连续区域高程测绘的潜力和能力仍未被充分探索。本研究开发了一种基于随机森林模型的空间连续大规模地形高程估算方法，利用ICESat-2、Sentinel-1和Sentinel-2的开放卫星数据。核心原理是构建一个随机森林模型，能够描述ICESat-2 ATL03地形高程与其对应的Sentinel-1和Sentinel-2的极化特征及光谱变量之间的复杂关系。通过整合这三颗不同卫星数据的优点，提出的方法能够在没有任何现场数据或手动设定参数的情况下，提供具有分米级垂直精度和10米空间/水平分辨率的地形高程估算。该方法在中国第三大岛（上海的崇明岛）的2021至2022年的高程数据上进行了测试。估算的地形高程与航空LiDAR衍生的高程进行了局部验证。此外，还与ICESat-2 ATL08的地形最佳拟合数据和全球生态系统动态调查L2A最低模式数据进行了全球视角的比较。预测的高程在两个航空LiDAR验证区域与测量高程之间表现出很高的相关性，均方根误差（RMSE）分别为0.34米和0.59米。不同地貌下预测高程的平均RMSE与ATL08和GEDI L2A衍生数据相比，分别为1.26米和1.18米。没有观察到显著的异常预测高程。这一发现表明，所提出的方法在不同地貌下表现出令人满意的鲁棒性，并且预测高程与整个岛屿的实际地形具有较好的一致性。我们所知，本研究首次基于新获得的卫星主动和被动遥感数据，无需任何地面真值调查、手动干预和先验知识，以10米分辨率进行高程测绘。与现有仅针对单一地貌的地形高程测绘研究不同，所提出的方法展示了ICESat-2在任何地貌和地覆盖类型下的能力和有效性，并显示出在区域/国家/全球尺度上进行高精度和高分辨率时间序列地形高程估算和更新的巨大潜力。
【doi】
https://doi.org/10.1016/j.jag.2024.104010
【作者信息】
Siqi Yao, 华东师范大学，国家级江河口海岸研究中心，上海 200241，中国
Kai Tan, 华东师范大学，国家级江河口海岸研究中心，上海 200241，中国
Yanjun Wang, 湖南科技大学，湖南省测绘与遥感地理信息工程重点实验室，湘潭 411201，中国
Weiguo Zhang, 国家重点实验室：河口与海岸研究，华东师范大学，上海 200241，中国
Shuai Liu, 国家重点实验室：河口与海岸研究，华东师范大学，上海 200241，中国
Jianru Yang，国家重点实验室：河口与海岸研究，华东师范大学，上海 200241，中国
论文63
Multilevel intuitive attention neural network for airborne LiDAR point cloud semantic segmentation
多层次直观注意力神经网络用于航空LiDAR点云语义分割
【摘要】
Three-dimensional laser scanning technology is widely employed in various fields due to its advantage in rapid acquisition of geographic scene structures. Achieving high precision and automated semantic segmentation of three-dimensional point cloud data remains a vital challenge in point cloud recognition. This study introduces a Multilevel Intuitive Attention Network (MIA-Net) designed for point cloud segmentation. MIA-Net consists of three key components: local trigonometric function encoding, feature sampling, and intuitive attention interaction. Initially, trigonometric encoding captures fine-grained local semantics within disordered point clouds. Subsequently, a multilayer perceptron addresses point-cloud feature pyramid construction, and feature sampling is performed using the point offset mechanism in the different levels. Finally, the multilevel intuitive attention(MIA) mechanism facilitates feature interactions across different layers, enabling the capture of both local attention features and global structure. The point-offset attention scheme introduced in this study significantly reduces computational complexity compared to traditional attention mechanisms, enhancing computational efficiency while preserving the advantages of attention mechanisms. To evaluate the results of MIA-Net, the ISPRS Vaihingen benchmark, LASDU and GML airborne datasets were tested. Experiments show that our network can achieve state-of-art performance in terms of Overall Accuracy(OA) and average F1-score(e.g., reaching 96.2% and 66.7% for GML datasets, respectively).
【摘要翻译】
三维激光扫描技术因其在快速获取地理场景结构方面的优势而被广泛应用于各个领域。然而，实现高精度和自动化的三维点云数据语义分割仍然是点云识别中的一个重要挑战。本研究介绍了一种用于点云分割的多层直观注意力网络（MIA-Net）。MIA-Net包含三个关键组件：局部三角函数编码、特征采样和直观注意力交互。首先，三角函数编码捕捉了无序点云中的细粒度局部语义。接下来，多层感知机处理点云特征金字塔构建，并在不同层级中使用点偏移机制进行特征采样。最后，多层直观注意力（MIA）机制促进了不同层次间的特征交互，使得局部注意力特征和全局结构能够被捕捉。研究中引入的点偏移注意力方案显著降低了与传统注意力机制相比的计算复杂度，提高了计算效率，同时保留了注意力机制的优势。为了评估MIA-Net的结果，我们在ISPRS Vaihingen基准测试、LASDU和GML空中数据集上进行了测试。实验表明，我们的网络在整体精度（OA）和平均F1得分方面可以实现最先进的性能（例如，GML数据集上的OA达到96.2%和F1得分达到66.7%）。
【doi】
https://doi.org/10.1016/j.jag.2024.104020
【作者信息】
Ziyang Wang, 虚拟地理环境重点实验室（南京师范大学）；教育部，南京 210023，中国；江苏省地理信息资源开发与应用协同创新中心，南京210023，中国；南京师范大学地理科学学院，南京，中国
Hui Chen, 南京大学地理与海洋科学学院，南京 210023，中国
Jing Liu, 虚拟地理环境教育部重点实验室（南京师范大学），江苏省地理信息资源开发与应用协同创新中心，南京师范大学地理学院，南京 210023，中国
Jiarui Qin, 虚拟地理环境重点实验室（南京师范大学），教育部；江苏省地理信息资源开发与应用协同创新中心；南京师范大学地理学院，南京 210023，中国
Yehua Sheng, 虚拟地理环境重点实验室（南京师范大学），教育部；江苏省地理信息资源开发与应用协同创新中心；南京师范大学地理学院，南京 210023，中国
Lin Yang，南京师范大学虚拟地理环境重点实验室，教育部，南京 210023，中国；江苏省地理信息资源开发与应用协同创新中心，南京210023，中国；南京师范大学地理学院，南京，中国
论文64
A novel bathymetric signal extraction method for photon-counting LiDAR data based on adaptive rotating ellipse and curve iterative fitting
基于自适应旋转椭圆和曲线迭代拟合的新型光子计数LiDAR数据水深信号提取方法
【摘要】
Equipped with the Advanced Topographic Laser Altimeter System (ATLAS) of 532 nm, the Ice, Cloud and Land Elevation Satellite-2 (ICESat-2) is enabled to penetrate the water surface to retrieve water depth information at certain depths. However, signal photon extraction is significantly impacted by noise and varying terrain conditions, particularly in deep water regions with weaker signal photons. To address these challenges, this study proposed a novel bathymetric signal extraction method for ICESat-2/ATLAS data based on adaptive rotating ellipse and B-spline curve iterative filtering. First, raw photons are segregated into water surface and bottom photons using the specular return removal algorithm and combined with the RANdom SAmple Consensus (RANSAC) algorithm to derive water surface elevation. Second, our method employs the modified Ordering Points to Identify the Clustering Structure (OPTICS) with adaptive variable ellipse and B-spline curve iterative filtering to detect bottom signal photons. Following refraction and tidal corrections, water depth is computed. Finally, the bathymetric accuracy of the proposed algorithm is evaluated using manually labeled photons and airborne bathymetric LiDAR data. The experimental results indicate the proposed algorithm’s superior F_score value, which increased by about 6.6 % and 4.1 % compared with the high-confidence photons and Adaptive Variable Ellipse Filtering Bathymetric Method (AVEBM). Moreover, the Mean Absolute Deviation (MAE) of bathymetric accuracy is 0.47 m, the Root Mean Square Error (RMSE) is 0.55 m, and Coefficient of Determination (R2) of bathymetric accuracy is 0.93. The proposed method effectively extracts signal photons and provides accurate water depth for nearshore bathymetry estimation.
【摘要翻译】
配备532 nm的高级地形激光高度计系统（ATLAS），冰云陆地高程卫星-2（ICESat-2）能够穿透水面，获取一定深度的水深信息。然而，信号光子提取受到噪声和地形条件的显著影响，特别是在深水区域，信号光子较弱。为解决这些挑战，本研究提出了一种基于自适应旋转椭圆和B样条曲线迭代滤波的新型水深信号提取方法。首先，利用镜面反射去除算法将原始光子分离为水面光子和底部光子，并结合随机样本一致性算法（RANSAC）导出水面高程。其次，我们的方法采用改进的有序点识别聚类结构（OPTICS）算法，通过自适应变量椭圆和B样条曲线迭代滤波来检测底部信号光子。在经过折射和潮汐校正后，计算水深。最后，使用人工标记光子和航空水深激光雷达数据评估了该算法的水深精度。实验结果表明，所提出算法的F_score值优于高置信光子和自适应变量椭圆滤波水深方法（AVEBM），分别提高了约6.6%和4.1%。此外，该算法的水深精度平均绝对误差（MAE）为0.47米，均方根误差（RMSE）为0.55米，水深精度的决定系数（R2）为0.93。该方法有效地提取了信号光子，并为近岸水深估计提供了准确的水深数据。
【doi】
https://doi.org/10.1016/j.jag.2024.104042
【作者信息】
Zijia Wang, 数字地球科学重点实验室，中国科学院航天信息研究所，北京 100094，中国；国际大数据可持续发展目标研究中心，北京 100094，中国；中国科学院大学资源与环境学院，北京 100049，中国
Sheng Nie, 数字地球科学重点实验室，中国科学院航天信息研究所，北京 100094，中国；国际大数据可持续发展目标研究中心，北京 100094，中国
Cheng Wang, 数字地球科学重点实验室，中国科学院航天信息研究所，北京 100094，中国；国际大数据可持续发展目标研究中心，北京 100094，中国；资源与环境学院，中国科学院大学，北京 100049，中国
Bihong Fu, 数字地球科学重点实验室，中国科学院航天信息研究所，北京 100094，中国；国际大数据可持续发展目标研究中心，北京 100094，中国
Xiaohuan Xi, 数字地球科学重点实验室，中国科学院航天信息研究所，北京 100094，中国；国际大数据可持续发展目标研究中心，北京 100094，中国
Bisheng Yang，国家测绘工程信息工程重点实验室，武汉大学，武汉 430079，中国
论文65
ColorMesh: Surface and texture reconstruction of large-scale scenes from unstructured colorful point clouds with adaptive automatic viewpoint selection
ColorMesh：从无结构的彩色点云中重建大规模场景的表面和纹理，并进行自适应自动视角选择
【摘要】
High-quality surface reconstruction and texture reconstruction of large-scale scene play a pivotal role in the domains of ancient architecture, cultural heritage preservation, and 3D urban modeling. Given that point cloud data has emerged as a crucial medium for representing three-dimensional spatial information, its utilization for surface and texture reconstruction becomes indispensable. In this research, we propose a novel framework for reconstructing surface and texture from unstructured colorful point clouds without normal information to restore large-scale real-world scenes. Specifically, we first introduce an automatic virtual viewpoint selection method to generate virtual views by rendering the point cloud from multiple viewpoints. Subsequently, we construct a two-step network to facilitate accurate visibility prediction and texture inpainting for each virtual view. Then, the visibility information from multiple perspectives is integrated to solve an optimization problem incorporating visibility constraints, resulting in the generation of a 3D mesh. Subsequently, texture information from various perspectives is integrated, filtering techniques are applied to determine the optimal perspective for texturing, and a texture atlas is generated. Precise texture mapping is then performed, ultimately leading to the production of a comprehensive textured model. In contrast to alternative learning-based methodologies, our framework exclusively learns from two-dimensional images, encompassing the prediction of both visible and invisible points as well as the execution of image inpainting tasks. This approach exhibits exceptional versatility in managing large-scale point clouds while effectively leveraging the color and intensity attributes of the data for precise texture reconstruction. The experimental results demonstrate that our approach achieves a significant improvement of 2.06% in F-scores for outdoor surface reconstruction compared to the current state-of-the-art learning-based methods, while also outperforming them in texture reconstruction.
【摘要翻译】
高质量的大规模场景表面重建和纹理重建在古代建筑、文化遗产保护以及三维城市建模等领域中扮演着关键角色。由于点云数据已成为表示三维空间信息的重要媒介，其在表面和纹理重建中的应用变得不可或缺。在这项研究中，我们提出了一种新的框架，用于从无法提供法线信息的无结构彩色点云中重建表面和纹理，以恢复大规模现实世界场景。具体而言，我们首先介绍了一种自动虚拟视点选择方法，通过从多个视点渲染点云来生成虚拟视图。随后，我们构建了一个两步网络，便于准确的可见性预测和每个虚拟视图的纹理修复。然后，将来自多个视角的可见性信息整合起来，解决包含可见性约束的优化问题，生成三维网格。接着，整合各个视角的纹理信息，应用过滤技术以确定最佳的纹理视角，并生成纹理图集。然后，进行精确的纹理映射，最终生成一个全面的纹理模型。与其他基于学习的方法相比，我们的框架仅从二维图像中学习，包括可见和不可见点的预测以及图像修复任务。该方法在处理大规模点云时表现出卓越的通用性，同时有效利用数据的颜色和强度属性进行精确的纹理重建。实验结果表明，我们的方法在户外表面重建方面比目前最先进的学习方法提高了2.06%的F值，同时在纹理重建方面也表现优越。
【doi】
https://doi.org/10.1016/j.jag.2024.104041
【作者信息】
Mubai Li, 教育部3D信息获取与应用重点实验室，北京师范大学，北京 100048，中国；资源环境与旅游学院，北京师范大学，北京 100048，中国
Zhenxin Zhang, 教育部3D信息获取与应用重点实验室，北京师范大学，北京 100048，中国；资源环境与旅游学院，北京师范大学，北京 100048，中国
Siyun Chen, 教育部3D信息获取与应用重点实验室，北京师范大学，北京 100048，中国；资源环境与旅游学院，北京师范大学，北京 100048，中国
Liqiang Zhang, 国家遥感科学重点实验室，地理科学学院，北京师范大学，北京 100875，中国
Zhihua Xu, 中国矿业大学（北京）地球科学与测绘工程学院，北京 100083，中国
Xiaoxu Ren, 北京师范大学三维信息获取与应用教育部重点实验室，北京 100048，中国；北京师范大学资源环境与旅游学院，北京 100048，中国
Jinlan Liu, 北京师范大学三维信息获取与应用教育部重点实验室，北京 100048，中国；北京师范大学资源环境与旅游学院，北京 100048，中国
Peng Sun，中国北京市 100048，首都师范大学资源环境与旅游学院
论文66
A review and future directions of techniques for extracting powerlines and pylons from LiDAR point clouds
从LiDAR点云中提取电力线和电塔技术的综述与未来方向
【摘要】
The rapid progression of the intelligent grid requires continuous vigilance in monitoring and maintaining extensive powerline corridors to ensure their safety. In this context, LiDAR technology, renowned for its exceptional precision and reduced vulnerability to external interference, emerges as a valuable alternative for monitoring powerline corridors. This contrasts with conventional methods such as manual field inspections and imprecise sensors. However, the vast amount of data generated by LiDAR presents significant challenges, including scene noise, diverse scenarios, and unwanted objects proximate to powerlines or pylons. These factors complicate the accurate extraction and analysis of relevant data from point clouds produced by LiDAR. This review examines recent methodologies aimed at overcoming these challenges. It begins with a brief exploration of data collection systems for powerline corridors, including TLS, MLS, UAVLS, ALS, and CIR, highlighting their respective merits and drawbacks. The subsequent sections of the review provide a comprehensive overview of three methodological categories: tracking and detection-based approaches, machine learning-based techniques, and deep learning-based methods. Within each category, representative techniques are delineated, elucidating their potential, limitations, and applicable domains. This review incorporates qualitative analysis to enhance researchers' comprehension of current studies and to providea nuanced understanding of the strengths and weaknesses of these techniques. In a departure from previous research, this review extends its focus beyond powerline extraction to include the extraction of pylons and single wires. It identifies a notable oversight in the lack of emphasis on individual wire extraction, attributing this to challenges posed by wire proximity, and highlights limited attention to pylon extraction near vegetation. While machine learning and deep learning methods offer heightened automation, persistent issues such as the requirement for extensive labeled samples and inadequate model generalization, underscore the need for continued efforts to address these challenges. This discussion emphasizes the necessity of overcoming these hurdles to boost ongoing advancements in powerline and pylon extraction techniques.
【摘要翻译】
智能电网的快速发展要求对广泛的输电线路走廊进行持续的监测和维护，以确保其安全。在这种背景下，LiDAR 技术因其卓越的精度和对外部干扰的较低敏感性，成为监测输电线路走廊的宝贵替代方案。这与传统的方法，如人工现场检查和不够精确的传感器形成对比。然而，LiDAR 生成的大量数据带来了重大挑战，包括场景噪声、多样的场景和靠近输电线路或电缆塔的杂物。这些因素使得从 LiDAR 产生的点云中准确提取和分析相关数据变得复杂。本文综述了旨在克服这些挑战的最新方法。首先，简要探讨了用于输电线路走廊的数据采集系统，包括 TLS、MLS、UAVLS、ALS 和 CIR，突出其各自的优缺点。接下来的部分全面概述了三类方法学：基于跟踪和检测的方法、基于机器学习的技术和基于深度学习的方法。在每一类中，描述了代表性技术，阐明了它们的潜力、局限性和适用领域。该综述通过定性分析增强了研究人员对当前研究的理解，并提供了对这些技术优缺点的细致了解。与以往研究不同，本文将关注点扩展到电缆塔和单根电缆的提取。指出了在电缆提取上缺乏重视的显著遗漏，将其归因于电缆之间的距离带来的挑战，并指出在植被附近电缆塔提取的关注不足。尽管机器学习和深度学习方法提供了更高的自动化，但持续的问题，如对大量标注样本的需求和模型泛化不足，突显了需要继续努力解决这些挑战。本讨论强调了克服这些障碍的必要性，以推动输电线路和电缆塔提取技术的持续进展。
【doi】
https://doi.org/10.1016/j.jag.2024.104056
【作者信息】
Yueqian Shen, 中国南京 211100 河海大学地球科学与工程学院
Junjun Huang, 中国南京 211100 河海大学地球科学与工程学院
Jinguo Wang, 中国南京 211100 河海大学地球科学与工程学院
Jundi Jiang, 中国南京 211100 河海大学地球科学与工程学院
Junxi Li, 中国南京 211100 河海大学地球科学与工程学院
Vagner Ferreira，中国南京 211100 河海大学地球科学与工程学院
论文67
Satellite remote sensing and bathymetry co-driven deep neural network for coral reef shallow water benthic habitat classification
卫星遥感与水深测量共同驱动的深度神经网络用于珊瑚礁浅水底栖栖息地分类
【摘要】
Shallow-water benthic habitat classification of coral reefs based on satellite remote sensing is an important part of coral reef monitoring. Leveraging its potent capacity for feature learning, and generalization, deep learning emerges as a robust method for coral reef benthic habitat classification. Due to the complexity of the marine environment, it is difficult to produce high-quality pixel-by-pixel labels for deep learning-based methods, which makes it challenging to recover structural details of coral reef benthic habitats. Bathymetry data can provide spatial contextual information and geometric features, serving as auxiliary features to provide abundant structural information for benthic classification models. However, how to use bathymetry and what kind of bathymetry features to employ for assisting model learning remains to be explored. Therefore, a bathymetry feature fusion-weakly supervised coral reef benthic habitat classification model (BFFBHCM) is proposed. BFFBHCM is supervised by sparse benthic habitat samples with bathymetry and can generate dense, multi-scale bathymetry features. With the robust bathymetry-benthic feature fusion module (B-BFFM), BFFBHCM can consider both semantic and structural details of the benthic habitats, thus generating highly accurate benthic habitat classification results. Experiments were conducted using the NJUReef + dataset containing 10 coral reefs in the Spratly Islands, China, constructed based on in-situ data. Comprehensive experimental results demonstrate that the proposed BFFBHCM is insensitive to the vertical error in bathymetry, with an average mIoU 22.54 % higher than state-of-the-art methods. Furthermore, it outperforms the weakly-supervised method that excludes bathymetry by 10.14 %, and still exhibits generalization to coral reefs in different regions around the world.
【摘要翻译】
基于卫星遥感的浅水底栖栖息地分类是珊瑚礁监测的重要部分。由于其强大的特征学习和泛化能力，深度学习成为珊瑚礁底栖栖息地分类的一种有效方法。然而，由于海洋环境的复杂性，生成高质量的逐像素标签对深度学习方法而言非常困难，这使得恢复珊瑚礁底栖栖息地的结构细节具有挑战性。水深数据可以提供空间上下文信息和几何特征，作为辅助特征为底栖分类模型提供丰富的结构信息。然而，如何使用水深数据以及使用什么类型的水深特征来辅助模型学习仍需探索。因此，提出了一种水深特征融合弱监督珊瑚礁底栖栖息地分类模型（BFFBHCM）。BFFBHCM通过稀疏的底栖栖息地样本和水深数据进行监督，并能够生成密集的多尺度水深特征。通过强大的水深-底栖特征融合模块（B-BFFM），BFFBHCM能够同时考虑底栖栖息地的语义和结构细节，从而生成高度准确的底栖栖息地分类结果。实验使用了包含中国南沙群岛10个珊瑚礁的NJUReef +数据集，该数据集基于实地数据构建。综合实验结果表明，所提出的BFFBHCM对水深的垂直误差不敏感，平均mIoU比最先进的方法高22.54%。此外，相较于排除水深数据的弱监督方法，其性能提高了10.14%，并且在全球不同区域的珊瑚礁上仍表现出良好的泛化能力。
【doi】
https://doi.org/10.1016/j.jag.2024.104054
【作者信息】
Hui Chen, 江苏省地理信息科学与技术重点实验室（南京大学），南京 210023，中国；南京大学地理与海洋科学学院，南京 210023，中国；南海研究协同创新中心（南京大学），南京 210023，中国；新型软件技术与产业化协同创新中心（南京大学），南京210023，中国
Jian Cheng, 江苏省地理信息科学与技术重点实验室（南京大学），南京 210023，中国；南京大学地理与海洋科学学院，仙林路 163 号，南京 210023，中国；南海研究协同创新中心（南京大学），仙林路 163 号，南京 210023，中国；新型软件技术与产业化协同创新中心（南京大学），南京210023，中国
Xiaoguang Ruan, 浙江水利水电学院测绘学院，中国杭州市 310018
Jizhe Li, 江苏省地理信息科学与技术重点实验室（南京大学），中国南京 210023；南京大学地理与海洋科学学院，中国南京 210023，仙林路 163 号；南海研究协同创新中心（南京大学），中国南京 210023，仙林路 163 号；新型软件技术与产业化协同创新中心（南京大学），中国南京210023
Li Ye, 江苏省地理信息科学与技术重点实验室（南京大学），中国南京 210023；南京大学地理与海洋科学学院，仙林路 163 号，中国南京 210023；南海研究协同创新中心（南京大学），仙林路 163 号，中国南京 210023；新型软件技术与产业化协同创新中心（南京大学），中国南京 210023
Sensen Chu, 江苏省地理信息科学与技术重点实验室（南京大学），中国南京 210023；南京大学地理与海洋科学学院，仙林路 163 号，中国南京 210023；南海研究协同创新中心（南京大学），仙林路 163 号，中国南京 210023；新型软件技术与产业化协同创新中心（南京大学），中国南京210023
Liang Cheng, 江苏省地理信息科学与技术重点实验室（南京大学），中国南京 210023；南京大学地理与海洋科学学院，仙林路 163 号，中国南京 210023；南海研究协同创新中心（南京大学），仙林路 163 号，中国南京 210023；新型软件技术与产业化协同创新中心（南京大学），中国南京210023
Ka Zhang，南京师范大学虚拟地理环境重点实验室（教育部），中国南京 210023；江苏省地理信息资源开发与应用协同创新中心，中国南京 210023
论文68
Statistical characteristics of merger-type sea-breeze fronts and associated circulation patterns in the Bohai Bay region, North China
渤海湾地区合并型海风锋面及相关环流模式的统计特征
【摘要】
Sea breeze front (SBF) is one of the important weather systems affecting the occurrence and development of severe convective weather in the Bohai Bay region (BBR). 226 cases of merger-type SBFs (MSBFs) merged with gust fronts (GFs) and convective systems (CSs), respectively, were identified based on Doppler weather radar data and ground-based automatic weather station data from May to September during 2009–2018 in the BBR, and their basic tempo-spatial characteristics and associated atmospheric circulation backgrounds are documented for the first time.The number of MSBFs cases merged with GFs (MSBF-GFs) and that of MSBFs merged with CSs (MSBF-CSs) were 172 and 54, respectively. The number of MSBFs varied significantly in each year, with 37 (13) in the most (least) frequent year, and with an average number of 22.6 per year. More than 93.8 % of the MSBFs occurred from June to August, especially most frequent (37.2 %) in July. The merging locations of the MSBFs were mainly distributed in the central-northern Tianjin and the southeastern Hebei province, and the horizontal scales of MSBFs were mainly distributed in the range of 130–309 km. About 29.6 % (51.9 %) of the MSBF-CSs cases resulted in significantly (slightly) enhanced convections, while 51.4 % (23.8 %) of the MSBF-GFs bring about significantly (slightly) enhanced convections. About 72.1 % of the MSBF-GFs cases are merged in near “face-to-face” form, and their 49.2 % (23.4 %) proportion lead to significantly (slightly) enhanced convections. The atmospheric circulation patterns of MSBFs identified using objective classification method showed that, the major two patterns (occupied 56.6 % cases) have similar dynamic, thermodynamic, and water vapor characteristics including westerlies or southwesterlies with intensity about 8–10 m/s at 500 hPa, showing significant warm and moist air delivered from the south and relatively weak vertical wind shear along with intense water vapor convergence at 850 hPa.
【摘要翻译】
海风锋（SBF）是影响渤海湾地区（BBR）严重对流天气发生和发展的重要天气系统之一。基于2009年至2018年5月至9月的多普勒天气雷达数据和地面自动气象站数据，识别出了226个海风锋合并型（MSBFs）与阵风锋（GFs）和对流系统（CSs）的合并情况，并首次记录了它们的基本时空特征及相关的大气环流背景。MSBF与GFs（MSBF-GFs）的案例有172个，而与CSs（MSBF-CSs）的案例有54个。每年的MSBF数量差异显著，最多年份为37个，最少年份为13个，年均数量为22.6个。超过93.8%的MSBF发生在6月至8月之间，尤其在7月最为频繁，占37.2%。MSBF的合并位置主要分布在天津中北部和河北省东南部，MSBF的水平尺度主要集中在130至309公里范围内。约29.6%（51.9%）的MSBF-CSs案例导致显著（轻微）增强的对流，而51.4%（23.8%）的MSBF-GFs案例导致显著（轻微）增强的对流。约72.1%的MSBF-GFs案例以接近“面对面”的形式合并，其中49.2%（23.4%）的比例导致显著（轻微）增强的对流。利用客观分类方法识别的MSBF的大气环流模式显示，主要有两种模式（占56.6%的案例），具有类似的动力学、热力学和水汽特征，包括500 hPa层的西风或西南风，强度约为8-10 m/s，显示南方输送的显著暖湿空气和850 hPa层相对较弱的垂直风切变，以及强烈的水汽收敛。
【doi】
https://doi.org/10.1016/j.jag.2024.104005
【作者信息】
Abuduwaili Abulikemu, 新疆绿洲生态重点实验室，新疆大学地理与遥感科学学院，乌鲁木齐，中国
Zhiyi Li, 新疆绿洲生态学重点实验室；新疆大学地理与遥感科学学院，中国乌鲁木齐
Jingjing Zheng, 乌鲁木齐市城市岩土工程勘察测绘研究所；乌鲁木齐基础地理信息中心，中国乌鲁木齐
Shushi Zhang, 中国气象局交通气象重点实验室，南京气象联合研究所，南京，中国；中国气象科学研究院重大天气国家重点实验室，北京，中国；中国气象局高影响天气重点实验室，长沙，中国
Xin Xu, 中国气象局/南京大学中尺度重型天气重点实验室和大气科学学院，南京，中国；国家重点天气实验室和中国气象局/南京大学联合大气雷达研究中心，北京，中国
Yan Wang, 天津市海洋气象重点实验室，天津市天气改造办公室，天津市气象局，中国气象局，天津，中国
Yiwei Liu，天津市海洋气象重点实验室，天津市天气改造办公室，天津市气象局，中国气象局，天津，中国
论文69
Monitoring and evaluation of the effects of Grain for Green Project on the Loess Plateau: A case study of Wuqi County in China
对黄土高原“退耕还林”项目效果的监测与评估：以中国吴起县为例
【摘要】
The Loess Plateau plays a significant role in the implementation of China’s Grain for Green Project due to severe ecological damage in the region. In order to monitor and evaluate the effects of Grain for Green Project, a study was conducted in Wuqi County, which is representative of the Loess Plateau. The study utilized remote sensing (RS) and geographic information system (GIS) technologies to analyze the spatial and temporal patterns of Grain for Green Project and assess its effects. The findings indicate that the Grain for Green Project resulted in notable improvements in Wuqi County from 2000 to 2018. Firstly, there was a significant increase in vegetation coverage, accompanied by a reduction in soil erosion intensity. Secondly, approximately 64 % of cropland was converted, leading to an expansion of forest and grassland areas. Thirdly, the focus of vegetation restoration was primarily on converting cropland to grassland, indicating its suitability for the county compared to forestation. Lastly, the conversion of steep cropland (>25°) was influenced by the density of less steep cropland (<25°). This study emphasizes the importance of guiding farmers in selecting appropriate vegetation restoration strategies and finding a balance between erosion control and agricultural production within the Grain for Green Project. Furthermore, the study recognizes that the project’s significant effects are not solely attributed to land use conversion but also to the self-restoration of vegetation. This shift towards a self-restoration perspective is crucial for the future high-quality development of the Grain for Green Project.
【摘要翻译】
黄土高原在中国“退耕还林还草”项目的实施中扮演了重要角色，因为该地区的生态损害严重。为了监测和评估该项目的效果，我们在代表黄土高原的吴起县进行了研究。研究利用遥感（RS）和地理信息系统（GIS）技术分析了“退耕还林还草”项目的时空模式，并评估了其效果。研究结果表明，从2000年到2018年，“退耕还林还草”项目在吴起县取得了显著改善。首先，植被覆盖度显著增加，土壤侵蚀强度降低。其次，大约64%的耕地被转换，森林和草地面积扩展。第三，植被恢复的重点主要是将耕地转换为草地，显示出其相对于造林更适合该县的情况。最后，陡坡耕地（>25°）的转换受到了相对平缓耕地（<25°）密度的影响。本研究强调了指导农民选择合适的植被恢复策略以及在“退耕还林还草”项目中平衡侵蚀控制与农业生产的重要性。此外，研究认识到该项目的显著效果不仅归因于土地使用转换，还包括植被的自我恢复。这种向自我恢复视角的转变对未来“退耕还林还草”项目的高质量发展至关重要。
【doi】
https://doi.org/10.1016/j.jag.2024.104006
【作者信息】
Ying Liu, 中国科学院与教育部土壤水分保持与生态环境研究中心，陕西省杨凌市 712100，中国；中国科学院与水利部水土保持研究所，陕西省杨凌市712100，中国；中国科学院大学，北京市 100049，中国
Chenxiao Kong, 中国科学院与教育部土壤水分保持与生态环境研究中心，陕西省杨凌市 712100，中国；中国科学院与水利部水土保持研究所，陕西省杨凌市712100，中国；中国科学院大学，北京市 100049，中国
Yueni Zhang, 西北农林科技大学林学院，陕西省杨凌市 712100，中国
Guan Liu, 西北农林科技大学林学院，陕西省杨凌市 712100，中国
Jinghua Huang, 中国科学院和教育部土壤与水保持与生态环境研究中心，陕西杨凌 712100，中国；中国科学院和水利部土壤与水保持研究所，陕西杨凌712100，中国
Guoqing Li, 中国科学院与教育部土壤与水保持与生态环境研究中心，陕西杨凌 712100，中国；中国科学院与水利部土壤与水保持研究所，陕西杨凌712100，中国
Sheng Du，中国科学院与教育部土壤与水保持与生态环境研究中心，陕西杨凌 712100，中国；中国科学院与水利部土壤与水保持研究所，陕西杨凌712100，中国
论文70
Rapid and extensive expansion of shrub encroachment into grassland in Xilin Gol League, China, and its driving forces
中国锡林郭勒盟灌木侵占草原的快速和广泛扩展及其驱动因素
【摘要】
Shrub encroaching on grasslands threatens grassland ecosystems and negatively affects human land use and livelihood. However, existing research on this subject is limited, having focused on small areas, thus hindering a comprehensive understanding of the diverse patterns of large-scale shrub invasions into grasslands. In recent decades, a significant encroachment of shrubs into Xilin Gol grassland has been observed in China, resulting in considerable damage to the grasslands. Therefore, this study aimed to investigate the rapid expansion of shrub-encroached grassland (SEG) from 1990 to 2020 in the Xilin Gol League, Inner Mongolia Autonomous Region, China, using long-term Landsat images. The spatial distribution characteristics of the new SEG areas were analyzed. The main factors driving the shrub encroachment in the study area were identified using a geographic detector. Our results revealed a continuous expansion of SEG in the study area over the past three decades, marked by an increase of 48.76 × 103 km2 and an average annual gain rate of 9.41 %. The increased SEG was primarily owing to grassland transition. The sprawl of the SEG was the most prominent in the western and northern regions of the study area, extending to gentle slopes of approximately 5° and low-middle elevations between 600–1300 m above mean sea level. Shrub encroachment was primarily influenced by surface temperature, wind speed, and relative humidity, with elevation, slope, and precipitation exerting relatively weaker influences. Rather than being driven by a singular factor, the expansion was a result of the combined influence of various factors. This study provides a valuable case study for understanding shrub invasion dynamics in arid and semi-arid regions globally.
【摘要翻译】
灌木侵入草原对草原生态系统构成威胁，并对人类土地使用和生计产生负面影响。然而，现有的研究较为有限，主要集中在小范围内，阻碍了对大规模灌木侵入草原多样化模式的全面理解。近年来，中国锡林郭勒草原出现了显著的灌木侵入现象，对草原造成了严重损害。因此，本研究旨在利用长期的Landsat影像，调查1990年至2020年间内蒙古自治区锡林郭勒盟灌木侵入草原（SEG）的快速扩展情况，并分析新形成的SEG区域的空间分布特征。研究还使用地理探测器确定了驱动研究区域灌木侵入的主要因素。我们的结果表明，在过去三十年中，研究区域内的SEG持续扩展，面积增加了48.76 × 10^3 km²，年均增长率为9.41%。SEG的增加主要归因于草原的转变。SEG的扩展在研究区域的西部和北部最为显著，扩展至大约5°的缓坡以及海拔600-1300米的低中海拔地区。灌木侵入主要受到地表温度、风速和相对湿度的影响，而海拔、坡度和降水的影响相对较弱。扩展并非由单一因素驱动，而是多种因素综合影响的结果。本研究为理解全球干旱和半干旱地区的灌木侵入动态提供了有价值的案例。
【doi】
https://doi.org/10.1016/j.jag.2024.104009
【作者信息】
Xiaoqing Lv, 中国陕西省西安市710127，西北大学城市与环境科学学院
Jianhong Liu, 中国陕西省西安市710127，西北大学城市与环境科学学院；英国莱斯特大学，环境未来研究所，景观与气候研究中心，地理、地质与环境学院，空间公园莱斯特，92 Corporation Road，莱斯特 LE4 5SP
Heiko Balzter, 莱斯特大学，环境未来研究所，景观与气候研究中心，地理、地质与环境学院，空间公园莱斯特，92号公司路，莱斯特 LE4 5SP，英国；国家地球观测中心，莱斯特大学，空间公园莱斯特，92号公司路，莱斯特 LE4 5SP，英国
Ziyue Dong, 西北大学，城市与环境科学学院，陕西省西安市 710127，中国
Jinnuo Li, 中国西安710127，西北大学城市与环境科学学院
Wei Zhang, 中国西安710127，西北大学城市与环境科学学院
Yige Guo，中国西安710054，西安科技大学地质与环境学院
论文71
Impacts of droughts and human activities on water quantity and quality: Remote sensing observations of Lake Qadisiyah, Iraq
干旱和人类活动对水量和水质的影响：伊拉克卡迪西亚湖的遥感观测
【摘要】
Water quantity and quality in lakes are closely linked to the compounding effects of climate change and human activities in their catchments, especially for lakes located in semi-arid and arid regions where water resources are scarce. Whilst knowledge gaps exist for these effects in semi-arid and arid region lakes due mainly to the lack of long-term in situ monitoring data. By using satellite remote sensing data, this study firstly investigated the variations of water level, chlorophyll-a concentration (Chl-a) and turbidity in Lake Qadisiyah, Iraq between 2000 and 2019. Results showed that the average water level was 138.3 m in 2000–2019, it decreased clearly in 2001, 2009, 2015 and 2018 with the lowest value of 120 m in July 2015. The mean Chl-a was 6.3 mg/m3 and it showed an overall increasing trend during 2000 and 2019. Turbidity showed extremely high values (>10 NTU) in 2009 and 2017–2018 compared to the mean value of 3.6 NTU in 2000–2019. The boosted regression tree (BRT) was then used to explore the relationship between those variations and El Niño-Southern Oscillation, droughts, meteorological factors and land use land cover changes in the catchment. Results revealed that water level declines were mainly associated with droughts led by La Niña events. Chl-a increase in the lake were mainly explained by built-up area increase and water area decrease in the catchment, with a relative contribution of 29.2 % and 28.6 % respectively. Water area changes in the catchment were the main factor influencing turbidity explaining 55.3 % of the variation. An exception water level decline in 2014–2016 was also observed when there was no drought, which was most likely caused by the cut off of water flow upstream and the release of water from the dam during periods of war. The findings in this study underscored the impacts of climate and human activities on water quantity and quality in semi-arid region lakes. Actions such as improving water use efficiency, establishing water storages, and enhancing cross-border cooperation are therefore recommended to deal with extreme events. Pollution control measures in the catchment are also suggested to prevent water quality deterioration in the lake.
【摘要翻译】
水体数量和质量在湖泊中与气候变化和人类活动的复合效应密切相关，尤其是对于位于半干旱和干旱地区的湖泊，这些地区水资源稀缺。然而，由于缺乏长期的原位监测数据，对这些效应的了解仍存在空白。通过使用卫星遥感数据，本研究首次调查了2000年至2019年间伊拉克卡迪西亚湖（Lake Qadisiyah）水位、叶绿素a浓度（Chl-a）和浊度的变化。结果表明，2000年至2019年间，平均水位为138.3米，但在2001年、2009年、2015年和2018年明显下降，其中2015年7月水位最低，为120米。叶绿素a的平均值为6.3毫克/立方米，并且在2000年至2019年间显示出总体上升趋势。与2000年至2019年的平均值3.6 NTU相比，2009年和2017年至2018年的浊度值极高（>10 NTU）。随后，使用提升回归树（BRT）探索了这些变化与厄尔尼诺-南方涛动、干旱、气象因素以及流域土地使用/覆盖变化之间的关系。结果揭示，水位下降主要与由拉尼娜事件引发的干旱有关。湖泊中叶绿素a的增加主要由流域内建成区增加和水域减少解释，相对贡献分别为29.2%和28.6%。流域内水域变化是影响浊度的主要因素，解释了55.3%的变化。2014年至2016年水位下降的例外情况也被观察到，当时没有干旱，最可能是由于战争期间上游水流被切断和水库释放水源所致。本研究的发现突显了气候和人类活动对半干旱地区湖泊水量和水质的影响。因此，建议采取提高水资源利用效率、建立水储存设施和增强跨境合作等措施来应对极端事件。同时，建议采取污染控制措施，以防止湖泊水质恶化。
【doi】
https://doi.org/10.1016/j.jag.2024.104021
【作者信息】
Dalin Jiang, 地球与行星观测科学（EPOS），生物与环境科学，自然科学学院，斯特灵大学，斯特灵 FK9 4LA，英国
Ian Jones, 地球与行星观测科学（EPOS），生物与环境科学，自然科学学院，斯特灵大学，斯特灵 FK9 4LA，英国
Xiaohan Liu, 普利茅斯海洋实验室，普利茅斯 PL1 3DH，英国
Stefan G.H. Simis, 普利茅斯海洋实验室，普利茅斯 PL1 3DH，英国
Jean-François Cretaux, LEGOS，图卢兹大学，法国国家航天研究中心（CNES），法国国家科学研究中心（CNRS），法国开发研究院（IRD），图卢兹大学，31400 图卢兹，法国
Clement Albergel, 欧洲航天局气候办公室，ECSAT，哈威尔园区，迪德科特，牛津郡，英国
Andrew Tyler, 地球与行星观测科学（EPOS），生物与环境科学，自然科学学院，斯特林大学，斯特林 FK9 4LA，英国
Evangelos Spyrakos，地球与行星观测科学（EPOS），生物与环境科学，自然科学学院，斯特林大学，斯特林 FK9 4LA，英国
论文72
High spatial resolution inversion of chromophoric dissolved organic matter (CDOM) concentrations in Ebinur Lake of arid Xinjiang, China: Implications for surface water quality monitoring
在中国新疆干旱区伊比努尔湖中高分辨率反演色度溶解有机物（CDOM）浓度：对地表水质量监测的意义
【摘要】
Utilizing satellite remote sensing for the assessment and temporal-spatial analysis of Chromophoric Dissolved Organic Matter (CDOM) is vital for overseeing lake water health and devising management plans. This study focused on the saline, turbid and arid Ebinur Lake, located in China’s northwestern region. The Random Forest (RF) and eXtreme Gradient Boosting (XGBoost) model algorithms were compared to select the one with the highest accuracy. It combined Sentinel-2 remote sensing data and in situ measurement data for the quantitative inversion of CDOM. Monthly CDOM distribution maps were generated with a 10 m resolution for the non-frozen months of May to October from 2018 to 2022, followed by a comprehensive analysis of temporal trends. The primary conclusions are: (1) The XGBoost model yielded highly accurate CDOM estimates, with a training set coefficient of determination (R2) of 0.94, a Root Mean Square Error (RMSE) of 0.06 mg/L, Mean Absolute Percentage Error (MAPE) of 6.05 %, Relative Percent Difference (RPD) of 4.07; the test set demonstrated an R2 of 0.41 with an RMSE of 0.22 mg/L, MAPE of 22.74 %, RPD of 1.35; (2) Throughout the study period, the main lake portion displayed variable CDOM spatial patterns and trends. The inversion indicated higher CDOM concentrations in the central part than nearshore areas and decreasing CDOM in tandem with seasonable water-surface shrinkage. The findings offer hints for an accurate evaluation of water color parameters of Ebinur Lake and practical references for monitoring arid-region lake water quality via remote sensing.
【摘要翻译】
利用卫星遥感技术评估和时空分析色度溶解有机物（CDOM）对于监测湖泊水质和制定管理计划至关重要。本研究以位于中国西北地区的咸水、浑浊和干旱的伊犁湖为研究对象，比较了随机森林（RF）和极端梯度提升（XGBoost）模型算法，以选择准确度最高的模型。研究结合了Sentinel-2遥感数据和现场测量数据对CDOM进行定量反演。生成了2018年至2022年5月至10月（非冻结月份）的月度CDOM分布图，分辨率为10米，并进行了全面的时效性分析。主要结论如下：(1) XGBoost模型提供了高度准确的CDOM估计，训练集的决定系数（R2）为0.94，均方根误差（RMSE）为0.06 mg/L，平均绝对百分比误差（MAPE）为6.05%，相对百分比差异（RPD）为4.07；测试集显示R2为0.41，RMSE为0.22 mg/L，MAPE为22.74%，RPD为1.35；(2) 在整个研究期间，湖泊主要区域表现出不同的CDOM空间模式和趋势。反演结果显示，中央区域的CDOM浓度高于近岸区域，并且随着季节性水面缩小CDOM浓度降低。这些发现为准确评估伊犁湖的水色参数提供了线索，并为通过遥感监测干旱地区湖泊水质提供了实用参考。
【doi】
https://doi.org/10.1016/j.jag.2024.104022
【作者信息】
Zhihui Li, 新疆大学地理与遥感科学学院，中国乌鲁木齐 830046
Cheng Chen, 浙江师范大学地理与环境科学学院，中国金华 321004
Naixin Cao, 新疆大学地理与遥感科学学院，中国乌鲁木齐 830046
Zhuohan Jiang, 新疆大学地理与遥感科学学院，中国乌鲁木齐 830046
Changjiang Liu, 新疆师范大学地理科学与旅游学院，中国乌鲁木齐 830054
Saheed Adeyinka Oke, 南非布隆方丹中央科技大学土木工程系，邮政编码9300
Chiyung Jim, 香港教育大学社会科学系，香港大埔罗屏路，香港，中国
Kaixuan Zheng, 浙江省海洋学院，杭州 310012，中国；海洋空间资源管理技术重点实验室，杭州 310012，中国
Fei Zhang，浙江师范大学地理与环境科学学院，金华 321004，中国
论文73
Geodetector model-based quantitative analysis of vegetation change characteristics and driving forces: A case study in the Yongding River basin in China
基于地质探测模型的植被变化特征和驱动因素定量分析：以中国永定河流域为例
【摘要】
Vegetation is one of the most crucial components of terrestrial ecosystems, and monitoring vegetation change as well as studying the factors that drive its formation provide significant guidance for restoring ecological biodiversity. The choice of driving indicators for vegetation change in previous studies has not been comprehensive enough, and particularly groundwater depth has not been considered. Therefore, 10 natural factors and 5 human factors were chosen for our study. We adopted the normalized difference vegetation index (NDVI) to measure vegetation growth. In this study, we utilized trend analysis, the Mann-Kendall test, and the Hurst index to investigate the spatiotemporal variance of NDVI in the YDRB. The geographical detector model (Geodetector) was employed to examine vegetation change attributed to human and natural variables. As a result of the study, we found that over the past 22 years, the NDVI in the basin increased from 0.62 to 0.70, with an increase of +0.0040/yr. Land use type is the most significant driver affecting NDVI changes. The interaction of two factors has a greater effect on vegetation change more than a single factor. The relationship between land use type and annual mean precipitation explained 34.5 % of the change in vegetation. Groundwater depth contributed 4.1 % to the explanation of vegetation change. Furthermore, we have determined the optimal range of specific variables conducive to vegetation growth. The results help us further understand the potential driving mechanism of vegetation cover change in the YDRB and provide a theoretical reference for relevant managers to formulate the ecological restoration measures in the basin.
【摘要翻译】
植被是陆地生态系统中最关键的组成部分之一，监测植被变化及研究其驱动因素对恢复生态生物多样性具有重要指导意义。以往研究中选择的植被变化驱动指标不够全面，尤其是地下水位未被考虑。因此，本研究选择了10个自然因素和5个人为因素。我们采用归一化植被指数（NDVI）来测量植被生长。通过趋势分析、Mann-Kendall检验和赫斯特指数，本研究调查了黄河流域NDVI的时空变异性。地理探测器模型（Geodetector）被用来检验植被变化归因于人为和自然变量的影响。研究结果表明，在过去22年中，流域内NDVI从0.62上升到0.70，年均增长0.0040。土地利用类型是影响NDVI变化的最重要因素。两个因素的交互作用对植被变化的影响大于单一因素。土地利用类型与年均降水量之间的关系解释了34.5%的植被变化。地下水深度对植被变化的解释贡献了4.1%。此外，我们还确定了有利于植被生长的具体变量的最佳范围。结果有助于进一步理解黄河流域植被覆盖变化的潜在驱动机制，并为相关管理者制定流域生态恢复措施提供理论参考。
【doi】
https://doi.org/10.1016/j.jag.2024.104027
【作者信息】
Yujing Guo, 北京师范大学水科学学院，中国北京市 100875
Lirong Cheng, 水科学学院，北京师范大学，北京 100875，中国
Aizhong Ding, 水科学学院，北京师范大学，北京 100875，中国
Yumin Yuan, 四川大学建筑与环境学院，四川 610000，中国
Zhengyan Li, 北京师范大学水科学学院，北京 100875，中国
YiZhe Hou, 北京金河水资源与水电建设集团有限公司，北京 102206，中国
Liangsuo Ren, 南宁师范大学地理科学与规划学院，南宁 530001，中国
Shurong Zhang，北京师范大学地理学部，北京 100875，中国
论文74
Future challenges of terrestrial water storage over the arid regions of Central Asia
中亚干旱地区陆地水储量未来挑战
【摘要】
Since the arid regions of Central Asia (ACA) are located in the interior of Eurasia, water resources play a vital role in the stability of its ecosystem and economic development. Based on the terrestrial water storage anomaly (TWSA) of the Gravity Recovery and Climate Experiment (GRACE), we analyze the observed characteristics of the TWSA over the ACA during 2003–2014. Results indicate that the terrestrial water storage (TWS) in the region showed an overall declining trend from 2003 to 2014, and the autumn TWS in this region is the smallest compared to other seasons and exhibits a strong decreasing trend at least −4.5 cm/decade. This means water resources over the ACA are scarcer and more vulnerable in autumn. The Distance between Indices of Simulation and Observation (DISO) method is employed to evaluate the performance of the sixth phase of the Coupled Model Intercomparison Project (CMIP6) models in simulating the autumn TWSA over the ACA. Compared with observational results, the autumn TWSA values captured by CMIP6 models are larger and the declining TWS trends are weaker. Using the optimal CMIP6 models, the statistical downscaling method constrains the projection results of autumn TWSA values over the ACA using the GRACE datasets. It shows autumn TWS will continue to decrease in most parts of the ACA in the future, and water scarcity will be the most severe in Tajikistan and southwestern Kazakhstan. Under SSP126, Tajikistan’s TWSA is projected to decrease by 11.0 cm in the long term. This study reveals the current situation and possible future changes in TWS over the ACA in autumn, providing references for water resource management and sustainable development policies in this area to avoid losses caused by water scarcity.
【摘要翻译】
由于中亚干旱地区（ACA）位于欧亚大陆的内部，水资源在该地区生态系统的稳定性和经济发展中发挥着至关重要的作用。基于重力恢复与气候实验（GRACE）的陆地水储量异常（TWSA），我们分析了2003年至2014年期间中亚干旱地区TWSA的观测特征。结果表明，该地区的陆地水储量（TWS）在2003年至2014年期间总体呈下降趋势，而秋季的TWS在各个季节中最小，并且呈现出至少为-4.5厘米/十年的显著下降趋势。这意味着在秋季，干旱区的水资源更加稀缺和脆弱。我们采用了模拟与观测指数距离（DISO）方法来评估第六阶段耦合模型比较计划（CMIP6）模型在模拟中亚干旱区秋季TWSA方面的表现。与观测结果相比，CMIP6模型捕捉到的秋季TWSA值较大，且TWS下降趋势较弱。使用最优的CMIP6模型，统计降尺度方法结合GRACE数据集对中亚干旱区秋季TWSA值的预测结果进行了约束。研究显示，未来大多数地区的秋季TWS将继续下降，其中塔吉克斯坦和哈萨克斯坦西南部的水资源短缺将最为严重。在SSP126情景下，塔吉克斯坦的TWSA预计长期将减少11.0厘米。这项研究揭示了中亚干旱区秋季TWS的现状及可能的未来变化，为该地区的水资源管理和可持续发展政策提供了参考，以避免因水资源短缺而造成的损失。
【doi】
https://doi.org/10.1016/j.jag.2024.104026
【作者信息】
Yuzhuo Peng, 海洋气象与气候变化研究中心 实验室：国家海洋环境科学重点实验室 学院：海洋与地球学院 学校：厦门大学 地址：中国厦门 361102
Hao Zhang, 新疆生态与地理研究所沙漠与绿洲生态国家重点实验室, 中国科学院, 乌鲁木齐, 中国；中国科学院中亚生态与环境研究中心, 乌鲁木齐, 中国；中国科学院大学, 北京, 中国
Zhuo Zhang, 新疆生态与地理研究所沙漠与绿洲生态国家重点实验室, 中国科学院, 乌鲁木齐, 中国；中国科学院中亚生态与环境研究中心, 乌鲁木齐, 中国；中国科学院大学, 北京, 中国
Bin Tang, 中国科学院大学, 北京, 中国；中国科学院大气物理研究所大气科学与地球物理流体动力学数值模拟国家重点实验室 (LASG), 北京, 中国
Dongdong Shen, 中国科学院大学, 北京, 中国；中国科学院大气物理研究所大气科学与地球物理流体动力学数值模拟国家重点实验室 (LASG), 北京, 中国
Gang Yin, 新疆大学地理与遥感科学学院, 乌鲁木齐, 中国；新疆大学绿洲生态学重点实验室, 乌鲁木齐, 中国
Yaoming Li, 中国科学院新疆生态与地理研究所沙漠与绿洲生态国家重点实验室, 乌鲁木齐, 中国；中国科学院中亚生态与环境研究中心, 乌鲁木齐, 中国；中国科学院大学, 北京, 中国，lym@ms.xjb.ac.cn
Xi Chen, 中国科学院新疆生态与地理研究所沙漠与绿洲生态国家重点实验室，乌鲁木齐，中国；中国科学院中亚生态与环境研究中心，乌鲁木齐，中国；中国科学院大学，北京，中国
Zengyun Hu, 中国热带疾病研究中心全球健康学院，上海交通大学医学院，上海 200025，中国；上海交通大学医学院公共卫生学院，上海 200025，中国
Sulaimon Habib Nazrollozoda，塔吉克斯坦农业科学院兽医研究所，卡霍罗夫街43号，734005 杜尚别，塔吉克斯坦
论文75
An attention-enhanced spatial–temporal high-resolution network for irrigated area mapping using multitemporal Sentinel-2 images
基于多时相Sentinel-2影像的注意力增强空间-时间高分辨率网络用于灌溉区映射
【摘要】
Accurate mapping of irrigated croplands is crucial for a comprehensive understanding of agricultural practices and land management. Despite recent advancements, there remains room for further exploration of the effective fusion of temporal information from multitemporal remote sensing images, which is essential for capturing the dynamic nature of agricultural landscapes. Many existing irrigation mapping methods concatenate multitemporal images in a direct way and thus neglect the temporal relationships within the image time series, especially the sequence and interdependencies of the temporal dimension. To address this gap, a novel deep learning model, named the attention-enhanced spatial–temporal high-resolution network (AEST-HRNet), which incorporates parallel processing and a fusion mechanism of multiresolution information streams, three-dimensional (3D) spatial–temporal convolution, and temporal attention modules, was proposed. When applied to irrigated regions in Washington and California, USA, AEST-HRNet effectively extracted irrigated areas using multitemporal Sentinel-2 images obtained with the Google Earth Engine (GEE). To validate the results, 208 representative sample patches were selected, and the AEST-HRNet maps were compared against third-party ground reference data and statistics from the United States National Agricultural Statistics Survey (NASS). Quantitative assessment revealed an impressive F1-score of 0.956, an intersection over union (IoU) value of 0.867, and an overall accuracy (OA) value of 0.973 in Washington, outperforming publicly released maps. Comparative evaluations demonstrated that AEST-HRNet outperforms pixel-based classification using the random forest (RF) model and convolution-based semantic segmentation methods based on metrics such as F1-score, IoU, and Kappa. This study introduces a promising solution for precise irrigation mapping, offering increased accuracy and efficiency in producing reliable irrigation maps.
【摘要翻译】
准确的灌溉耕地制图对全面了解农业实践和土地管理至关重要。尽管近年来取得了进展，但在有效融合多时相遥感图像的时间信息方面仍有进一步探索的空间，这对于捕捉农业景观的动态特性至关重要。许多现有的灌溉制图方法以直接方式拼接多时相图像，因此忽视了图像时间序列中的时间关系，特别是时间维度的序列和相互依赖关系。为了解决这一问题，提出了一种新型深度学习模型——注意力增强空间-时间高分辨率网络（AEST-HRNet），该模型结合了并行处理和多分辨率信息流的融合机制、三维（3D）空间-时间卷积以及时间注意力模块。应用于美国华盛顿州和加利福尼亚州的灌溉区时，AEST-HRNet有效地提取了使用Google Earth Engine（GEE）获取的多时相Sentinel-2图像的灌溉区域。为了验证结果，选择了208个代表性样本区块，将AEST-HRNet地图与第三方地面参考数据和美国国家农业统计调查（NASS）数据进行了比较。定量评估结果显示，华盛顿州的F1得分为0.956，交集与并集比值（IoU）为0.867，总体准确率（OA）为0.973，表现优于公开发布的地图。比较评估表明，AEST-HRNet在F1得分、IoU和Kappa等指标上优于基于随机森林（RF）模型的像素分类和基于卷积的语义分割方法。本研究提出了一种有前景的精确灌溉制图解决方案，提供了更高的准确性和效率，生成可靠的灌溉地图。
【doi】
https://doi.org/10.1016/j.jag.2024.104040
【作者信息】
Wei Li, 生态与环境学院，新疆大学，乌鲁木齐 830017，中国；新疆生态与地理研究所，中国科学院，乌鲁木齐 830011，中国；中国科学院大学，北京 100049，中国
Qinchuan Xin, 新疆生态与地理研究所，中国科学院，乌鲁木齐 830011，中国；中国科学院中亚生态与环境研究中心，乌鲁木齐 830011，中国；南方海洋科学与工程广东省实验室（珠海），中山大学地理与规划学院，广州 510275，中国
Ying Sun, 中山大学南方海洋科学与工程广东省实验室（珠海）地理与规划学院，广州 510275，中国
Yanqing Zhou, 新疆大学生态与环境学院，乌鲁木齐 830017，中国；中国科学院新疆生态与地理研究所，乌鲁木齐 830011，中国；中国科学院大学，北京 100049，中国
Jiangyue Li, 中国科学院新疆生态与地理研究所，乌鲁木齐 830011，中国；中国科学院中亚生态与环境研究中心，乌鲁木齐 830011，中国
Yidan Wang, 香港科技大学土木与环境工程系，中国香港
Yu Sun, 中国科学院新疆生态与地理研究所，新疆乌鲁木齐 830011，中国；中国科学院大学，北京 100049，中国
Guangyu Wang, 中国科学院新疆生态与地理研究所，新疆乌鲁木齐 830011，中国；中国科学院大学，北京 100049，中国
Ren Xu, 中国科学院新疆生态与地理研究所，新疆乌鲁木齐 830011，中国；中国科学院大学，北京 100049，中国
Lu Gong, 新疆大学生态与环境学院，新疆乌鲁木齐 830017，中国；教育部绿洲生态重点实验室，新疆乌鲁木齐 830017，中国
Yaoming Li，中国科学院新疆生态与地理研究所，中国乌鲁木齐 830011；中国科学院大学，北京 100049，中国；中国科学院中亚生态与环境研究中心，中国乌鲁木齐 830011
论文76
Assessing the impacts of temperature extremes on agriculture yield and projecting future extremes using machine learning and deep learning approaches with CMIP6 data
使用CMIP6数据的机器学习和深度学习方法评估温度极端对农业产量的影响及预测未来极端事件
【摘要】
Climate change, particularly extreme weather events, has significantly affected various sectors, including agriculture, human health, water resources, sea levels, and ecosystems. It is anticipated that the intensity, duration, and frequency of these extremes will escalate in the future. This study aims to discover the association between temperature extremes and agricultural yield and to project these extremes using machine learning (ML) and deep learning (DL) models with CMIP6 (Coupled Model Intercomparison Project Phase 6) data under two SSPs (Shared Socioeconomic Pathways). A bi-wavelet coherence technique is employed to investigate the association, providing detailed information in both the frequency and time domains for the period of 1980–2014. Various ML and DL models are trained and tested for the periods of 1985–2004 and 2005–2014, respectively, with gradient boosting machine chosen for projecting temperature extremes based on its superior performance. Mann-Kendall test is used for trend analysis in the projected temperature extremes. The results indicate strong negative and positive associations between TN10p (Cold nights) and TN90p (Warm nights), respectively, with wheat production. Additionally, there is a long-term negative association of CSDI (Cold Spell Duration Indicator) and strong positive association of WSDI (Warm Spell Duration Indicator) with rice yield. Projected results show an increase and decrease under SSP2-4.5 and SSP5-8.5, respectively, in DTR (Diurnal Temperature Range) at most stations. TN10p will increase in the future at most stations, with exceptions such as Muree station where it decreases during 2025–2049 and then increases under both SSPs. Projections show that TXn (annual or monthly minimum value of daily maximum temp) will increase in the future, with Muree station exhibiting the lowest value close to zero, while the average maximum value is around 20 °C at Khanpur station. Trend analysis reveals significantly increasing trend in TR20 (Tropical nights) and decreasing trend in CSDI in future durations under both SSPs. These findings hold implications for policymakers and stakeholders in various departments, including agriculture, health, and water resources management.
【摘要翻译】
气候变化，尤其是极端天气事件，已经显著影响了农业、人类健康、水资源、海平面和生态系统等各个领域。预计这些极端事件的强度、持续时间和频率将在未来上升。本研究旨在探讨温度极端与农业产量之间的关系，并利用机器学习（ML）和深度学习（DL）模型结合CMIP6（耦合模型比较项目第六阶段）数据，在两个共享社会经济路径（SSP）下对这些极端情况进行预测。研究采用了双小波相干技术来研究这种关联，为1980–2014年期间的频率和时间域提供了详细信息。各种ML和DL模型在1985–2004年和2005–2014年期间进行训练和测试，其中梯度提升机（Gradient Boosting Machine, GBM）因其优越的性能被选用于预测温度极端事件。使用Mann-Kendall检验进行趋势分析。结果表明，TN10p（寒冷夜晚）与小麦生产之间存在显著的负相关，而TN90p（温暖夜晚）与小麦生产之间存在显著的正相关。此外，CSDI（寒冷波时长指标）与稻米产量之间存在长期的负相关，而WSDI（温暖波时长指标）与稻米产量之间存在显著的正相关。预测结果显示，在SSP2-4.5下，大多数站点的日温差（DTR）将增加，而在SSP5-8.5下，大多数站点的DTR将减少。未来大多数站点的TN10p将增加，例外的是穆雷（Muree）站点，在2025–2049年期间TN10p减少，然后在两个SSP下再次增加。预测显示TXn（年或月的每日最高温度最低值）将增加，而穆雷站点的值接近零，而Khanpur站点的平均最高值约为20°C。趋势分析揭示，在未来的两个SSP下，TR20（热带夜晚）存在显著增加趋势，而CSDI存在显著减少趋势。这些发现对农业、健康和水资源管理等多个部门的政策制定者和利益相关者具有重要的参考意义。
【doi】
https://doi.org/10.1016/j.jag.2024.104071
【作者信息】
Firdos Khan, 自然科学学院，国家科学技术大学，44000 伊斯兰堡，巴基斯坦；水文学遥感实验室，空间与遥感研究中心，国立中央大学，桃园市中大路300号，中坜区，320317 台湾，中国；统计学研究所，阿尔卑斯-亚德里亚大学，大学街65-67号，9020 克拉根福，奥地利
Yuei-An Liou, 水文学遥感实验室，空间与遥感研究中心，国立中央大学，桃园市中大路300号，中坜区，320317 台湾，中国
Gunter Spöck, 统计学研究所，阿尔卑斯-阿德里亚大学，大学街65-67号，9020 克拉根福，奥地利
Xue Wang, 中国科学院地理科学与资源研究所，陆面模式与模拟重点实验室，北京，100101，中国
Shaukat Ali，全球变化影响研究中心（GCISC），联邦气候变化部，44000 伊斯兰堡，巴基斯坦；多伦多儿童医院，加拿大
论文77
Assessing the responsiveness of multiple microwave remote sensing vegetation optical depth indices to drought on crops in Midwest US
评估多种微波遥感植被光学深度指数对中西部美国作物干旱的响应性
【摘要】
Agricultural drought is a major natural disaster affecting biomass accumulation and causing food loss, exacerbated by the increasing frequency of flash droughts and compounded drought-heatwave events. Traditional optical remote sensing indices cannot directly represent the water content of vegetation, resulting in a limited understanding of crop response to drought. To address this gap, we investigated the responsiveness of microwave Vegetation Optical Depth (VOD) with four bands (L-, C-, X-, KU-) and four emerging VOD-derived products to drought conditions in crops in the Midwest US. These products include the normalized Difference Between Night and Day VOD (nVOD), Slope of the Regression of Day and Night VOD (σ), Standardized VOD Index (SVODI), and VOD to estimate Gross Primary Productivity (VOD2GPP). They employ different theoretical modeling approaches to crop growth and water use strategies. We comprehensively analyzed the trend, seasonality, and residual of VODs, using Leaf Area Index (LAI) for comparison, and further assessed the lagged and cumulative effects, quantified drought sensitivity, and captured responsiveness to cumulative drought using thresholds. The results showed a time lag in the response of VOD series to drought as indicated by the Standardized Precipitation Evapotranspiration Index (SPEI). VODs achieved faster responses and higher correlations compared to LAI. Among them, VOD_L exhibited the most statistically significant pixels (39.84%) and positive
with 96.81% of all pixels. For cumulative effects, VOD_L, VOD_C, VOD_X, VOD_KU, and SVODI were highly correlated in the early stages of droughts. We also found that crops in Iowa exhibited medium to high drought sensitivity (average values of 0.55 to 0.74), with the highest drought sensitivity calculated using the isohydricity indicator, σ. Based on threshold comparison, σ showed a timely response in the first month of drought (average of −0.62), whereas VOD_L and VOD_C performed best in the second month (both averaging −1.85), and VOD2GPP (−2.94) was the most responsive in the third month. Due to water use strategy, maize responded more quickly to the onset of drought compared to soybeans. Overall, the results demonstrated that VOD is promising for crop phenology and drought research. This study provides the first comprehensive investigation of the diverse capabilities of multiple VOD-based indices in drought monitoring across various timescales and croplands.
【摘要翻译】
农业干旱是影响生物量积累并导致粮食损失的主要自然灾害，干旱的频发和干旱-热浪复合事件使这一问题更加严重。传统的光学遥感指数无法直接代表植被的水分含量，导致对作物对干旱反应的理解有限。为了填补这一空白，我们研究了微波植被光学厚度（VOD）在四个频段（L、C、X、KU）及其衍生产品对美国中西部作物干旱条件的响应。这些衍生产品包括夜间和白天VOD的归一化差异（nVOD）、夜间和白天VOD回归斜率（σ）、标准化VOD指数（SVODI）以及估算总初级生产力的VOD（VOD2GPP）。它们采用不同的理论建模方法来分析作物生长和水分利用策略。我们综合分析了VOD的趋势、季节性和残差，使用叶面积指数（LAI）进行比较，并进一步评估了滞后和累积效应，量化了干旱敏感性，并通过阈值捕捉了对累积干旱的响应。结果显示，VOD系列对干旱的响应存在时间滞后，这一点通过标准化降水蒸散发指数（SPEI）得到了体现。与LAI相比，VOD的响应更快、相关性更高。其中，VOD_L在统计上表现出最多的显著像素（39.84%）和96.81%的正相关Rmax-lag。对于累积效应，VOD_L、VOD_C、VOD_X、VOD_KU和SVODI在干旱早期阶段高度相关。我们还发现，爱荷华州的作物对干旱表现出中到高的敏感性（平均值为0.55到0.74），其中使用等水分性指标σ计算的干旱敏感性最高。根据阈值比较，σ在干旱的第一个月表现出了及时的响应（平均值为−0.62），而VOD_L和VOD_C在第二个月表现最佳（均平均为−1.85），VOD2GPP（−2.94）在第三个月的响应最为显著。由于水分利用策略，玉米对干旱的响应比大豆更快。总体而言，结果表明VOD在作物物候学和干旱研究中具有广阔的前景。本研究首次全面调查了多种VOD基础指数在不同时间尺度和作物区干旱监测中的多样化能力。
【doi】
https://doi.org/10.1016/j.jag.2024.104072
【作者信息】
Junjun Cao, 湖北省地理过程分析与模拟重点实验室/中华师范大学城市与环境科学学院，中国武汉430079
Yi Luo, 湖北省地理过程分析与模拟重点实验室/中华师范大学城市与环境科学学院，中国武汉430079
Xiang Zhang, 中国地质大学 地理与信息工程学院， 地理信息系统国家工程研究中心， 湖北 武汉 430074；湖北罗佳实验室，中国 武汉 430079；水资源部水文气象灾害机理与预警重点实验室，南京信息工程大学，中国南京
Lei Fan, 西南大学地理科学学院， 重庆金佛山喀斯特生态系统国家观测研究站， 重庆 400715
Jianbin Tao, 湖北省地理过程分析与模拟重点实验室/中华师范大学城市与环境科学学院，中国武汉430079
Won-Ho Nam, 韩国安成国立大学国家农业用水研究中心 社会安全与系统工程学院、农业环境科学研究所
Chanyang Sur, 韩国安成国立大学国家农业用水研究中心 社会安全与系统工程学院、农业环境科学研究所；韩国安城国立大学国家农业水研究中心
Yuqi He, 重庆测绘研究院，中国 重庆401121
Aminjon Gulakhmadov, 塔吉克斯坦国家科学院水问题、水电和生态研究所，塔吉克斯坦杜尚别734042
Dev Niyogi，德克萨斯大学奥斯汀分校土木、建筑与环境工程系，美国奥斯汀 78712
论文78
Quantitative evaluation of the impact of band optimization methods on the accuracy of the hyperspectral metal element inversion models
带优化方法对高光谱金属元素反演模型精度的定量评价
【摘要】
To reduce the high redundancy of band information in hyperspectral data, various band optimization methods have been adopted, which could be divided into two types namely band extraction (e.g., principal component analysis, PCA) and band selection (e.g., Spearman correlation coefficient, SRC). However, the applicability and effectiveness of different band optimization methods were rarely reported in the literature. Therefore, based on the rock sample data of the Baixintan deposit, we compare the performance of two band optimization algorithms (principal component analysis−based band extraction and SRC-based band selection) in inverting metal elements (Cu, Fe, Ni, Cr, Mg) using the adaptive genetic algorithms-gradient boosting regression tree (AGA-GBRT) algorithm. Two band optimization methods have shown different effects in improving the accuracy of target metal elements. The five models with the highest accuracy in metal elements include LT-R-Cu, LT-PCA-Fe, ORI-PCA-Ni, SDT-PCA-Cr, and LT-PCAMg. In the established model, the inversion accuracy of the Cu element is the lowest, possibly due to the high variability of the data itself (coefficient of variation is 3.55). Fe and Ni highly correlated with Cu elements were used to indirectly invert Cu element. Compared with the direct inversion model, the accuracy of the indirect inversion model has increased by 11%. Overall, PCA is more effective than SRC in predicting the content of metal elements in rocks. The conclusion presented in this article provides a ground experimental basis and technical support for future optimization of hyperspectral bands and inversion methods of metal element content in rocks.
【摘要翻译】
为了减少高光谱数据中带信息的高度冗余，已经采用了多种波段优化方法，这些方法可以分为两类：波段提取（例如主成分分析，PCA）和波段选择（例如斯皮尔曼相关系数，SRC）。然而，不同波段优化方法的适用性和有效性在文献中很少报道。因此，基于白兴潭矿床的岩石样本数据，我们比较了两种波段优化算法（基于主成分分析的波段提取和基于SRC的波段选择）在使用自适应遗传算法-梯度提升回归树（AGA-GBRT）算法反演金属元素（Cu、Fe、Ni、Cr、Mg）时的性能。这两种波段优化方法在提高目标金属元素的准确性方面表现出不同的效果。金属元素中准确性最高的五个模型包括LT-R-Cu、LT-PCA-Fe、ORI-PCA-Ni、SDT-PCA-Cr和LT-PCA-Mg。在建立的模型中，Cu元素的反演准确性最低，这可能是由于数据本身的高变异性（变异系数为3.55）。Fe和Ni与Cu元素高度相关，用于间接反演Cu元素。与直接反演模型相比，间接反演模型的准确性提高了11%。总体而言，PCA在预测岩石中金属元素含量方面比SRC更有效。本文提出的结论为未来高光谱波段优化和岩石中金属元素含量反演方法的优化提供了实验基础和技术支持。
【doi】
https://doi.org/10.1016/j.jag.2024.104011
【作者信息】
Xiumei Ma, 新疆生态与地理研究所沙漠与绿洲生态国家重点实验室，中国科学院，乌鲁木齐 830011，中国；新疆矿产资源与数字地质重点实验室，乌鲁木齐 830011，中国；中国科学院新疆矿产资源研究中心，乌鲁木齐 830011，中国；中国科学院大学，北京 100049，中国；根特大学，比利时根特 9000；中比地理信息联合实验室，比利时根特 9000；中比地理信息联合实验室，乌鲁木齐 830011，中国
Jinlin Wang, 新疆生态与地理研究所沙漠与绿洲生态国家重点实验室，中国科学院，乌鲁木齐 830011，中国；中国科学院空间利用技术与工程中心，北京 100094，中国；新疆矿产资源与数字地质重点实验室，乌鲁木齐 830011，中国；中国科学院新疆矿产资源研究中心，乌鲁木齐 830011，中国；中国科学院大学，北京 100049，中国
Kefa Zhou, 中国科学院空间利用技术与工程中心，北京 100094，中国
Wenqiang Zhang, 新疆生态与地理研究所沙漠与绿洲生态国家重点实验室，中国科学院，乌鲁木齐 830011，中国；中国科学院大学，北京 100049，中国；根特大学，比利时根特 9000；中比地理信息联合实验室，比利时根特 9000；中比地理信息联合实验室，乌鲁木齐 830011，中国
Zhixin Zhang, 新疆生态与地理研究所沙漠与绿洲生态国家重点实验室，中国科学院，乌鲁木齐 830011，中国；新疆矿产资源与数字地质重点实验室，乌鲁木齐 830011，中国；中国科学院新疆矿产资源研究中心，乌鲁木齐 830011，中国；中国科学院大学，北京 100049，中国
Shuguang Zhou, 中国科学院新疆生态与地理研究所沙漠与绿洲生态国家重点实验室，乌鲁木齐 830011，中国；新疆矿产资源与数字地质重点实验室，乌鲁木齐 830011，中国；中国科学院新疆矿产资源研究中心，乌鲁木齐 830011，中国；中国科学院大学，北京 100049，中国
Yong Bai, 中国科学院新疆生态与地理研究所沙漠与绿洲生态国家重点实验室，乌鲁木齐 830011，中国；中国科学院空间利用技术与工程中心，北京 100094，中国；新疆矿产资源与数字地质重点实验室，乌鲁木齐 830011，中国；中国科学院新疆矿产资源研究中心，乌鲁木齐 830011，中国；中国科学院大学，北京 100049，中国
Philippe De Maeyer, 根特大学，比利时根特 9000；中比地理信息联合实验室，比利时根特 9000；中比地理信息联合实验室，乌鲁木齐 830011，中国
Tim Van de Voorde，根特大学，比利时根特 9000；中比地理信息联合实验室，比利时根特 9000；中比地理信息联合实验室，乌鲁木齐 830011，中国
论文79
A light CNN based on residual learning and background estimation for hyperspectral anomaly detection
基于残差学习和背景估计的轻量级CNN用于高光谱异常检测
【摘要】
Existing deep learning-based hyperspectral anomaly detection methods typically perform anomaly detection by reconstructing a clean background. However, for the deep networks, there are many parameters that need to be adjusted. To reduce parameters of network and improve the performance of anomaly detection, a light CNN based on residual learning and background estimation was proposed. Different from traditional methods, the proposed method could directly learn anomaly features rather than background features. First, during the training stage, a background estimation method based on non-central convolution kernels was used to obtain the pseudo-background. Second, to purify the pseudo-background, a pair down-sampling method and a joint loss that combines cross-approximation background loss and consistency loss were proposed. Third, the anomaly matrix was obtained by the difference between the hyperspectral image (HSI) and the pseudo-background. Fourth, a light CNN with three layers was proposed to extract features of the anomaly matrix. Finally, during the prediction stage, anomaly detection results were calculated from the predicted anomaly matrix obtained by light CNN through the Mahalanobis distance. Experiments were conducted with multiple metrics on five real-world datasets. Compared with eight state-of-the-art methods, the proposed method achieved the superior performance in both qualitative and quantitative evaluations.
【摘要翻译】
现有的基于深度学习的高光谱异常检测方法通常通过重建干净的背景来进行异常检测。然而，对于深度网络来说，有许多参数需要调整。为了减少网络的参数并提高异常检测的性能，提出了一种基于残差学习和背景估计的轻量级卷积神经网络（CNN）。与传统方法不同，所提出的方法可以直接学习异常特征，而不是背景特征。首先，在训练阶段，使用基于非中心卷积核的背景估计方法来获取伪背景。其次，为了净化伪背景，提出了一种配对下采样方法和一种结合了交叉逼近背景损失和一致性损失的联合损失函数。第三，通过高光谱图像（HSI）和伪背景之间的差异来获得异常矩阵。第四，提出了一种具有三层的轻量级CNN，用于提取异常矩阵的特征。最后，在预测阶段，通过马哈拉诺比斯距离从通过轻量级CNN获得的预测异常矩阵中计算异常检测结果。通过在五个真实世界的数据集上使用多个指标进行实验，与八种最新的方法相比，所提出的方法在定性和定量评估中都表现出优越的性能。
【doi】
https://doi.org/10.1016/j.jag.2024.104069
【作者信息】
Jiajia Zhang, 西安电子科技大学，西安太白南路2号，710071，中国；墨尔本大学，格拉顿街，帕克维尔，墨尔本，3010，澳大利亚
Pei Xiang, 西安电子科技大学，西安太白南路2号，710071，中国
Jin Shi, 西安电子科技大学，西安太白南路2号，710071，中国
Xiang Teng, 西安电子科技大学，西安太白南路2号，710071，中国
Dong Zhao, 无锡大学，无锡 214105，中国
Huixin Zhou, 西安电子科技大学，中国陕西省西安市太白南路2号，邮编710071
Huan Li, 西安电子科技大学，中国陕西省西安市太白南路2号，邮编710071
Jiangluqi Song，西安电子科技大学，中国陕西省西安市太白南路2号，邮编710071
论文80
An ensemble framework for explainable geospatial machine learning models
解释性地理空间机器学习模型的集成框架
【摘要】
Analyzing spatially varying effects is pivotal in geographic analysis. However, accurately capturing and interpreting this variability is challenging due to the increasing complexity and non-linearity of geospatial data. Recent advancements in integrating Geographically Weighted (GW) models with artificial intelligence (AI) methodologies offer novel approaches. However, these methods often focus on single algorithms and emphasize prediction over interpretability. The recent GeoShapley method integrates machine learning (ML) with Shapley values to explain the contribution of geographical features, advancing the combination of geospatial ML and explainable AI (XAI). Yet, it lacks exploration of the nonlinear interactions between geographical features and explanatory variables. Herein, an ensemble framework is proposed to merge local spatial weighting scheme with XAI and ML technologies to bridge this gap. Through tests on synthetic datasets and comparisons with GWR, MGWR, and GeoShapley, this framework is verified to enhance interpretability and predictive accuracy by elucidating spatial variability. Reproducibility is explored through the comparison of spatial weighting schemes and various ML models, emphasizing the necessity of model reproducibility to address model and parameter uncertainty. This framework works in both geographic regression and classification, offering a novel approach to understanding complex spatial phenomena.
【摘要翻译】
在地理分析中，分析空间变异效应至关重要。然而，由于地理空间数据的日益复杂性和非线性，准确捕捉和解释这种变异性具有挑战性。近年来，地理加权（GW）模型与人工智能（AI）方法的结合提供了新颖的研究途径。然而，这些方法通常侧重于单一算法，并且更强调预测性而非可解释性。最新的GeoShapley方法将机器学习（ML）与Shapley值结合起来，解释地理特征的贡献，推进了地理空间机器学习与可解释人工智能（XAI）的结合。然而，该方法未能探讨地理特征与解释变量之间的非线性交互作用。在此，提出了一种集成框架，将局部空间加权方案与XAI和ML技术相结合，以填补这一空白。通过对合成数据集的测试，并与GWR、MGWR和GeoShapley进行比较，该框架被验证在通过阐明空间变异性来增强解释性和预测准确性方面具有优势。通过比较空间加权方案和各种ML模型，探讨了再现性问题，强调了解决模型和参数不确定性所需的模型再现性。这一框架适用于地理回归和分类，提供了一种理解复杂空间现象的新方法。
【doi】
https://doi.org/10.1016/j.jag.2024.104036
【作者信息】
Lingbo Liu，哈佛大学地理分析中心，美国马萨诸塞州剑桥市，邮编02138；武汉大学城市设计学院，中国湖北省武汉市，邮编430072
论文81
The daily soil water content monitoring of cropland in irrigation area using Sentinel-2/3 spatio-temporal fusion and machine learning
使用Sentinel-2/3时空融合和机器学习对灌溉区农田的日常土壤水分监测
【摘要】
Understanding soil moisture dynamics is crucial for crop growth. The digital mapping of field soil moisture distribution provides valuable information for agricultural water management. The optical satellite data provides fine scale soil moisture information for a region. However, these data are greatly limited due to cloud contamination and revisit period. Despite the reported beneficial effects of spatiotemporal fusion methods, the accurate estimates of high-resolution soil moisture through spatiotemporal fusion data are still unclear, particularly when using Sentinel-2/3 fusion images. This study introduces a new soil moisture estimation framework integrating spatio-temporal spectral information from Sentinel-2/3 fusion images and machine learning algorithm,and thus provide spatiotemporally continuous soil moisture estimation. The framework includes four fusion methods (ESTARRFM, Fit-FC, FSDAF and STFMF) and four machine learning models (PLSR, SVM, RF and GBRT). The feasibility of the framework was validated in the Hetao Irrigation Area of Inner Mongolia, China. The results showed that the Sentinel-2/3 fused image generated by Fit-FC was visually the closest to the true image, followed by ESTARFM, FSDAF, and STFMF. The spatiotemporal fusion-machine learning estimation framework provided reliable estimation for multi-layer (0 ∼ 20, 20 ∼ 40 and 40 ∼ 60 cm) soil water in the irrigation area. The dense time series of soil water generated by the framework facilitated the detection of irrigation events in the irrigated farmland. Our findings highlighted the effectiveness of Sentinel-2/3 fused images in providing high-resolution continuous daily monitoring of farmland soil water on a large scale. These high spatial–temporal resolution time series are valuable for monitoring crop growth and water resource management, contributing to further expanding the application of satellite remote sensing in precision agriculture.
【摘要翻译】
理解土壤水分动态对作物生长至关重要。田间土壤水分分布的数字化映射为农业水资源管理提供了宝贵的信息。光学卫星数据为区域提供了精细尺度的土壤水分信息。然而，这些数据由于云污染和重访周期的限制而受到很大制约。尽管已有报道显示时空融合方法具有积极效果，但通过时空融合数据准确估算高分辨率土壤水分的能力仍不明确，特别是在使用Sentinel-2/3融合图像时。本研究引入了一个新的土壤水分估算框架，该框架整合了Sentinel-2/3融合图像的时空光谱信息和机器学习算法，从而提供时空连续的土壤水分估算。该框架包括四种融合方法（ESTARRFM、Fit-FC、FSDAF和STFMF）和四种机器学习模型（PLSR、SVM、RF和GBRT）。该框架在中国内蒙古河套灌区的可行性得到了验证。结果显示，由Fit-FC生成的Sentinel-2/3融合图像在视觉上最接近真实图像，其次是ESTARFM、FSDAF和STFMF。时空融合-机器学习估算框架为灌区的多层（0~20厘米、20~40厘米和40~60厘米）土壤水分提供了可靠的估算。该框架生成的密集土壤水分时间序列有助于检测灌溉农田中的灌溉事件。我们的研究结果突出了Sentinel-2/3融合图像在大范围内提供高分辨率、连续、每日监测农田土壤水分的有效性。这些具有高时空分辨率的时间序列对于监测作物生长和水资源管理具有重要价值，进一步推动了卫星遥感在精准农业中的应用。
【doi】
https://doi.org/10.1016/j.jag.2024.104081
【作者信息】
Ruiqi Du, 西北农林科技大学水利与建筑工程学院，中国陕西省杨凌市，邮编712100
Youzhen Xiang, 西北农林科技大学水利与建筑工程学院，中国陕西省杨凌市，邮编712100
Junying Chen, 西北农林科技大学水利与建筑工程学院，中国陕西省杨凌市，邮编712100
Xianghui Lu, 南昌工程学院水利与生态工程学院，中国江西省南昌市，邮编330029；
江西省鄱阳湖流域生态水利工程技术创新中心，中国江西省南昌市，邮编330029
Fucang Zhang, 西北农林科技大学水利与建筑工程学院，中国陕西省杨凌市，邮编712100
Zhitao Zhang, 西北农林科技大学水利与建筑工程学院，中国陕西省杨凌市，邮政编码712100
Baocheng Yang, 南昌工程学院水利与生态工程学院，中国江西省南昌市，邮政编码330029；江西省鄱阳湖流域生态水利工程技术创新中心，中国江西省南昌市，邮政编码330029
Zijun Tang, 西北农林科技大学水利与建筑工程学院，中国陕西省杨凌市，邮政编码712100
Xin Wang, 西北农林科技大学水利与建筑工程学院，中国陕西省杨凌市，邮政编码712100
Long Qian，西北农林科技大学水利与建筑工程学院，中国陕西省杨凌市，邮政编码712100
论文82
Clustered remote sensing target distribution detection aided by density-based spatial analysis
基于密度空间分析的集群遥感目标分布检测
【摘要】
Small target detection in remote sensing is integral to a range of applications, including smart city systems and emergency rescue operations. However, the challenges posed by weak features and complex backgrounds in remote sensing images have hindered the efficacy of detection. Current models tend to focus on identifying individual targets, resulting in algorithms with larger parameters, slower detection efficiency, and difficulty in striking a balance between false positives and negatives. Given that many tasks do not require precise target location, a more efficient approach involves swiftly predicting target areas with models involving fewer parameters. This paper introduces the concept of group target distribution detection, gathering targets with similar distances and semantic similarities for clustered detection. A Gaussian probability map, formed from target density, is used to train a probability prediction model. We propose a new metric for evaluating this innovative group target distribution detection paradigm and provide a comparative assessment with traditional single-object detection models. In experimental evaluation, our proposed DenseUGE network — employing ResNet34 and ResNet50 as its backbone — surpasses the best baseline method by 3.37% on the AI-TOD dataset using our metrics. Additionally, visualizations demonstrate the ability of our proposed methodology to effectively identify the concentrated distribution of small target groups.
【摘要翻译】
遥感中的小目标检测在智能城市系统和紧急救援操作等多个应用中至关重要。然而，遥感图像中的弱特征和复杂背景带来的挑战影响了检测效果。当前模型往往侧重于识别单个目标，这导致了参数量较大、检测效率较低，并且在假阳性和假阴性之间难以取得平衡。由于许多任务不需要精确的目标位置，因此一种更高效的方法是利用参数较少的模型快速预测目标区域。本文引入了群体目标分布检测的概念，将具有相似距离和语义相似性的目标聚集起来进行集群检测。使用由目标密度形成的高斯概率图来训练概率预测模型。我们提出了一种新的评估指标来评估这种创新的群体目标分布检测范式，并与传统的单目标检测模型进行了比较评估。在实验评估中，我们提出的DenseUGE网络（采用ResNet34和ResNet50作为主干网络）在AI-TOD数据集上的表现超越了最佳基线方法3.37%。此外，视觉化结果展示了我们提出的方法有效识别小目标群体的集中分布能力。
【doi】
https://doi.org/10.1016/j.jag.2024.104019
【作者信息】
Puti Yan, 北京航空航天大学电子与信息工程学院，中国北京市
Jixiang Zhao, 新疆大学软件学院，中国乌鲁木齐市
Runze Hou, 清华大学深圳研究生院（Tsinghua-Berkeley Shenzhen Institute），中国深圳市
Xuguang Duan, 清华大学计算机科学与技术系，中国北京市
Shaoxiong Cai, 北京航空航天大学电子信息工程学院，中国北京市
Xin Wang，清华大学计算机科学与技术系，中国北京市；清华大学北京国家信息科学技术研究中心，中国北京市
论文83
Remote sensing image pan-sharpening via Pixel difference enhance
通过像素差异增强的遥感图像全色锐化
【摘要】
Nowadays, embedding-based pan-sharpening networks aimed at fusing panchromatic (PAN) and multispectral (MS) images are abundant, yet their results still show spectral distortion and spatial fuzziness. In this paper, we design a multi-scale fusion structure to minimize the gap between the pan-sharpened image and the reference image progressively. Specifically, we proposed a method based on the scale difference between PAN and MS images, using a convolutional neural network embedding pixel difference enhanced module (PDEM) to obtain the pan-sharpened image and minimizing the losses from each scale. The network includes three scales, each scale contains the PDEM to generate the intermediate results until to the last scale which obtains the final pan-sharpened result. The designed PDEM extracts deep features from PAN and MS images, using different kernel sizes and receptive field scales to diversify the extracted information. Three-direction pixel difference convolutions (PDCs) are utilized to maintain and enhance the edge details of spatial information. The loss function is to sum up the mean square error and mean absolute error between the pan-sharpened image and the reference image at three scales. Extensive experiments suggest the proposed method outperforms state-of-the-art methods from visual and quantitative perspectives, and confirm the effectiveness of PDEM in extracting and retaining image information and edge enhancement. The high-level vision task experiments also show our method has good practical value for further applications.
【摘要翻译】
当前，基于嵌入的全色锐化网络旨在融合全色（PAN）和多光谱（MS）图像，但其结果仍显示出光谱失真和空间模糊。在本文中，我们设计了一种多尺度融合结构，以逐步减少全色锐化图像与参考图像之间的差距。具体而言，我们提出了一种基于PAN和MS图像之间尺度差异的方法，使用卷积神经网络嵌入像素差异增强模块（PDEM）来获取全色锐化图像，并最小化每个尺度的损失。网络包括三个尺度，每个尺度包含PDEM以生成中间结果，直到最后一个尺度获得最终的全色锐化结果。设计的PDEM从PAN和MS图像中提取深层特征，使用不同的卷积核大小和感受野尺度来多样化提取的信息。三方向像素差异卷积（PDCs）用于保持和增强空间信息的边缘细节。损失函数是对三个尺度上全色锐化图像与参考图像之间的均方误差和均绝对误差进行求和。广泛的实验表明，所提出的方法在视觉和定量方面均优于现有的最先进方法，并确认了PDEM在提取和保留图像信息及边缘增强方面的有效性。高层视觉任务实验还表明，我们的方法在进一步应用中具有良好的实际价值。
【doi】
https://doi.org/10.1016/j.jag.2024.104045
【作者信息】
Xiaoxiao Feng, 湖北工业大学计算机学院，中国湖北省武汉市，邮政编码430068
Jiaming Wang, 武汉工程大学计算机科学与工程学院，中国湖北省武汉市，邮政编码430205
Zhiqi Zhang, 湖北工业大学计算机学院，中国湖北省武汉市，邮政编码430068
Xueli Chang，湖北工业大学计算机学院，中国湖北省武汉市，邮政编码430068
论文84
Precipitation nowcasting using transformer-based generative models and transfer learning for improved disaster preparedness
使用基于变换器的生成模型和迁移学习进行降水实时预报，以改善灾害准备
【摘要】
Due to the rapidly changing climate conditions, precipitation nowcasting poses a daunting challenge because it is impossible to make accurate short-term forecasts due to the rapid fluctuations in weather conditions. There are limitations to traditional methods of forecasting precipitation, such as the use of numerical models and radar extrapolation, when it comes to providing highly detailed and timely forecasts. With the help of contemporary machine learning (ML) models, including deep neural networks, transformers and generative models, complex precipitation nowcasting tasks can be performed in an efficient way. To address this critical task and enhance proactive emergency disaster management, we propose an innovative method based on transformer-based generative models for precipitation nowcasting. Our study area is the Soyang Dam basin in South Korea, located upstream of the Han River, characterized by a monsoon climate with approximately 1200 mm of annual precipitation. To develop a precipitation nowcasting model, radar composite data from 10 weather radars across South Korea is used. By utilizing radar reflective data in order to train our model, we are able to effectively predict future precipitation patterns, thus mitigating the risk of catastrophic weather conditions caused by heavy rainfalls. This dataset covers reflectivity data from 2018 to 2022, with a spatial resolution of 1km over a 960 × 1200 grid. Normalization using the min–max scaler method is applied to this reflectivity data, which is then transformed into grayscale images for uniform comparison. We enhance performance effectively by employing transfer learning with pre-trained Transformer models. Initially, we train the model using a comprehensive dataset. Subsequently, we fine-tune it for precipitation nowcasting using radar reflective data. This adaptation improves the accuracy of rainfall forecasting by capturing crucial features. Leveraging prior task knowledge through transfer learning not only enhances prediction accuracy but also increases overall efficiency. In terms of predictive accuracy, extensive experimental results demonstrate that our transformer-based nowcasting model outperforms related approaches, including conditional generative adversarial networks (cGANs), U-Net, convolutional long short-term memory (ConvLSTM), pySTEP. As a result of this research, disaster preparedness and response will be greatly improved through improved weather prediction.
【摘要翻译】
由于气候条件的迅速变化，降水短时预报面临巨大挑战，因为天气条件的快速波动使得难以进行准确的短期预测。传统的降水预报方法，如数值模型和雷达外推，在提供高度详细和及时的预报方面存在局限性。借助当代机器学习（ML）模型，包括深度神经网络、Transformer和生成模型，可以有效地执行复杂的降水短时预报任务。为应对这一关键任务并增强主动的灾害应急管理，我们提出了一种基于Transformer生成模型的创新降水短时预报方法。我们的研究区域为韩国汉江上游的昭阳江水库流域，该地区属于季风气候，年降水量约为1200毫米。为了开发降水短时预报模型，我们使用了韩国10个气象雷达的雷达复合数据。通过利用雷达反射数据来训练我们的模型，我们能够有效预测未来的降水模式，从而降低暴雨引发灾害性天气条件的风险。该数据集涵盖了2018年至2022年的反射率数据，空间分辨率为1公里，覆盖960×1200的网格。我们使用min-max标准化方法对这些反射率数据进行归一化处理，并将其转换为灰度图像以便于统一比较。我们通过使用预训练的Transformer模型进行迁移学习来有效地提升性能。首先，我们使用综合数据集训练模型；随后，我们使用雷达反射数据对其进行微调，以适应降水短时预报。这种调整通过捕捉关键特征提高了降水预测的准确性。通过迁移学习利用先前任务的知识，不仅提高了预测准确性，还提升了整体效率。在预测准确性方面，大量实验结果表明，我们的基于Transformer的短时预报模型优于相关方法，包括条件生成对抗网络（cGANs）、U-Net、卷积长短时记忆网络（ConvLSTM）和pySTEP。该研究的结果将大大提高天气预报的精度，从而显著改善灾害防范和应对能力。
【doi】
https://doi.org/10.1016/j.jag.2024.103962
【作者信息】
Md. Jalil Piran, 韩国首尔世宗大学计算机科学系，首尔，邮编05006
Xiaoding Wang, 福建师范大学计算机与网络安全学院，中国福建，邮编350117
Ho Jun Kim, 韩国首尔世宗大学计算机科学系，首尔，邮编05006
Hyun Han Kwon，韩国首尔世宗大学计算机科学系，首尔，邮编05006
论文85
A novel machine learning-based framework to extract the urban flood susceptible regions
基于机器学习的新框架提取城市洪水易发区域
【摘要】
The frequent occurrence of urban floods (UFs) poses significant threats to citizens’ lives and the national economy. Utilizing machine learning to assess urban flood susceptibility (UFS) provides valuable decision support for UF management. However, the precision of current studies is usually influenced by the variability of temporal factors like extreme rainfall, which limits the accurate identification of urban flood-susceptible regions (UFSRs). To address this issue, we present a novel approach that leverages the spatiotemporal distribution and characteristics of UFS to accurately identify UFSRs. In our case study of the Greater Bay Area (GBA) in China, we employed the Random Forest to assess the spatiotemporal distribution of UFS. We then used the Savitzky-Golay filter to correct UFS data based on the UFS time series from 2011 to 2020. The Theil-Sen median slope, Mann-Kendall test, and Hurst analysis were used to explore the spatiotemporal patterns of UFS. Shapley additive explanation was applied to quantify the contribution of selected variables. Our findings include: (1) UFS in the GBA demonstrates a rising trend, with high susceptibility areas increasing from 6.3 % in 2011 to 7.4 % in 2020; (2) UFSRs, covering approximately 11 % of the GBA, are primarily concentrated in the cities located around the central GBA; and (3) human behavior factors have a more significant influence on UF than natural ones. We believe the presented framework for the accurate extraction of UFSRs provides valuable decision support for sustainable city development.
【摘要翻译】
城市洪水的频繁发生对居民生活和国家经济构成了重大威胁。利用机器学习评估城市洪水易发性（UFS）为洪水管理提供了宝贵的决策支持。然而，目前研究的精度通常受到极端降雨等时间因素变化的影响，这限制了城市洪水易发区域（UFSRs）的准确识别。为了解决这一问题，我们提出了一种新方法，利用UFS的时空分布和特征来准确识别UFSRs。在中国粤港澳大湾区（GBA）的案例研究中，我们采用随机森林算法评估了UFS的时空分布。然后，我们基于2011年至2020年的UFS时间序列，使用Savitzky-Golay滤波器对UFS数据进行了校正。采用Theil-Sen中位斜率、Mann-Kendall检验和Hurst分析方法来探索UFS的时空模式。Shapley加法解释用于量化所选变量的贡献。我们的研究发现包括：（1）GBA的UFS呈上升趋势，高易发区从2011年的6.3%增加到2020年的7.4%；（2）覆盖GBA约11%的UFSRs主要集中在GBA中部周围的城市；（3）与自然因素相比，人为行为因素对城市洪水的影响更显著。我们相信，所提出的UFSRs准确提取框架为可持续城市发展提供了宝贵的决策支持。
【doi】
https://doi.org/10.1016/j.jag.2024.104050
【作者信息】
Xianzhe Tang, 广东省土地利用与整治重点实验室，华南农业大学，广州 510642，中国；自然资源与环境学院，环境与教育联合研究院，华南农业大学，广州 510642，中国
Juwei Tian, 广东省土地利用与整治重点实验室，华南农业大学，广州 510642，中国
Xi Huang, 广东省土地利用与整治重点实验室，华南农业大学，广州 510642，中国
Yuqin Shu, 华南师范大学地理科学学院，广州 510631，中国
Zhenhua Liu, 广东省土地利用与整治重点实验室，华南农业大学，广州 510642，中国
Shaoqiu Long, 广东省土地利用与整治重点实验室，华南农业大学，广州 510642，中国
Weixing Xue, 广东省土地利用与整治重点实验室，华南农业大学，广州 510642，中国
Luo Liu, 广东省土地利用与整治重点实验室，华南农业大学，广州 510642，中国
Xueming Lin, 自然资源与环境学院，环境与教育联合研究院，华南农业大学，广州 510642，中国
Wei Liu，国家资源环境信息系统重点实验室，地理科学与资源研究所，中国科学院，北京 100101，中国；中国科学院大学，北京 100049，中国
论文86
BiAU-Net: Wildfire burnt area mapping using bi-temporal Sentinel-2 imagery and U-Net with attention mechanism
BiAU-Net：使用双时相Sentinel-2影像和具有注意力机制的U-Net进行野火烧毁面积映射
【摘要】
The fusion of remote sensing and artificial intelligence, particularly deep learning, offers substantial opportunities for developing innovative methods in rapid disaster mapping and damage assessment. However, current models for wildfire burnt area detection and mapping are mostly constrained in handling imbalanced training samples with non-burnt areas oversampled, boundary areas with a mix of burnt and unburnt pixels, and regions with varying environmental contexts, leading to poor model generalizability. In response, this paper proposes a novel U-Net based model, known as BiAU-Net, which incorporates attention mechanisms and a well-designed loss function, enabling the model to focus on burnt areas and improve accuracy and efficiency, especially in detecting edges and small areas. Unlike traditional single-input U-Net models for image segmentation, the proposed BiAU-Net considers and incorporates temporal changes with two inputs, pre- and post-fire Sentinel-2 imagery, enhancing performance across diverse environmental areas. Five independent areas from different continents are selected as study cases, one for training the model and all five for testing, to demonstrate the generalizability of the proposed model. We used the Fire Disturbance Climate Change Initiative v5.1 product from the European Space Agency as a baseline for model evaluation. The experiment results indicate that BiAU-Net: (1) significantly outperformed the baseline with improvements of 11.56% in Overall Accuracy, 29.08% in Precision, 7.06% in Recall, 19.90% in F1-score, 15.44% in Balanced Accuracy, 29.90% in Kappa Coefficient, and 28.29% in Matthews Correlation Coefficient (MCC); (2) largely surpassed the performance of U-Net and its variants in most study areas; (3) demonstrated good generalizability in five testing areas across different continents; and (4) achieved the highest overall performance compared to the most state-of-the-art wildfire burnt area detection models, evidenced by the highest F1-score and MCC values.
【摘要翻译】
遥感与人工智能，特别是深度学习的融合，为快速灾害制图和损害评估方法的创新提供了巨大的机会。然而，目前的野火燃烧面积检测和制图模型主要在处理样本不平衡的问题上受到限制，特别是未燃烧区域被过度采样，边界区域存在燃烧与未燃烧像素混合，以及环境背景变化的区域，导致模型泛化能力较差。对此，本文提出了一种新型的基于U-Net的模型，称为BiAU-Net，该模型融合了注意力机制和精心设计的损失函数，使模型能够集中于燃烧区域，提高检测精度和效率，特别是在检测边界和小面积区域方面。与传统的单输入U-Net图像分割模型不同，BiAU-Net考虑并整合了时间变化，使用前后火灾的Sentinel-2影像作为两个输入，从而在不同环境区域中提升了性能。选取了来自不同大洲的五个独立区域作为研究案例，一个用于训练模型，其余五个用于测试，以展示模型的泛化能力。我们使用了欧洲空间局的火灾干扰气候变化计划v5.1产品作为模型评估的基线。实验结果表明，BiAU-Net: (1) 显著优于基线模型，整体准确度提高了11.56%，精确度提高了29.08%，召回率提高了7.06%，F1分数提高了19.90%，平衡准确度提高了15.44%，Kappa系数提高了29.90%，Matthews相关系数（MCC）提高了28.29%； (2) 在大多数研究区域中大幅超越了U-Net及其变体的表现； (3) 在五个不同大洲的测试区域中展示了良好的泛化能力； (4) 与最先进的野火燃烧面积检测模型相比，取得了最高的整体性能，F1分数和MCC值均为最高。
【doi】
https://doi.org/10.1016/j.jag.2024.104034
【作者信息】
Tang Sui, 美国威斯康星大学麦迪逊分校地理系，550 N Park St.，麦迪逊，WI 53706，美国；同济大学土地测绘与地理信息学院，上海，中国
Qunying Huang, 美国威斯康星大学麦迪逊分校地理系，550 N Park St.，麦迪逊，WI 53706，美国
Mingda Wu, 地理系，威斯康星大学麦迪逊分校，550 N Park St., Madison, 53706, WI, 美国；地球、气候与环境系，北伊利诺伊大学，DeKalb, 60115, IL, 美国
Meiliu Wu, 地理系，威斯康星大学麦迪逊分校，550 N Park St., Madison, 53706, WI, 美国
Zhou Zhang，生物系统工程系，威斯康星大学麦迪逊分校，460 Henry Mall, Madison, 53706, WI, 美国
论文87
Nonlinear effects of urban multidimensional characteristics on daytime and nighttime land surface temperature in highly urbanized regions: A case study in Beijing, China
城市多维特征对高密度城市化地区白天和夜间地表温度的非线性影响：以中国北京为例
【摘要】
It is crucial to clarify the nonlinear effects of urban multidimensional characteristics on land surface temperature (LST). However, the combined consideration of the urban green space (UGS), water bodies, buildings, and socio-economic factors is limited. And the diurnal differences in their thermal effects have been less considered. In this study, central Beijing was taken as study area. Local climate zones (LCZ) were firstly applied to reveal spatiotemporal heterogeneity of LST. Then, the interpretable machine learning methods were utilized to quantitatively reveal nonlinear thermal effects of urban multidimensional characteristics, i.e., the UGS, water bodies, and building landscape features, and socio-economic features. The results indicated that built type LCZs have a higher average LST compared to natural type LCZs. And the LST of built type LCZs is simultaneously influenced by buildings’ density and height characteristics. Daytime LST is mainly affected by the landscape proportions of UGS, buildings, and trees, while nighttime LST is more influenced by socio-economic and building characteristics. The thermal effects of key factors exhibit nonlinear characteristics. Whether during the day or night, the impact of building coverage on LST is greater than that of building height, consistently exhibiting a warming effect. While, the building height and water body edge density factors both exhibited a reversal trend in their thermal impact between day and night. Our study also emphasized the importance of trees type in UGS and provided recommendations for UGS planning based on sensitivity and contribution considerations. These findings can help to regulate urban LST and promote sustainable urban development.
【摘要翻译】
澄清城市多维特征对地表温度（LST）非线性效应至关重要。然而，结合考虑城市绿地（UGS）、水体、建筑物和社会经济因素的研究还较为有限。尤其是它们的昼夜热效应差异较少被关注。本研究以北京市中心为研究区域。首先应用局部气候区（LCZ）来揭示LST的时空异质性。然后，使用可解释的机器学习方法定量揭示城市多维特征（即UGS、水体、建筑景观特征和社会经济特征）的非线性热效应。结果表明，与自然型LCZ相比，建成型LCZ具有更高的平均LST。而建成型LCZ的LST同时受到建筑密度和高度特征的影响。白天的LST主要受到UGS、建筑物和树木景观比例的影响，而夜间的LST则更受社会经济和建筑特征的影响。关键因素的热效应展现出非线性特征。无论白天还是夜晚，建筑覆盖率对LST的影响大于建筑高度，持续展现出加热效应。而建筑高度和水体边缘密度因素在昼夜之间的热影响则表现出反转趋势。我们的研究还强调了UGS中树木类型的重要性，并基于敏感性和贡献考虑提出了UGS规划建议。这些发现有助于调控城市LST，促进可持续城市发展。
【doi】
https://doi.org/10.1016/j.jag.2024.104067
【作者信息】
Wenxiu Liu, 航空航天信息研究院，中国科学院，北京 100094，中国；中国科学院大学，北京 100049，中国
Linlin Zhang, 航空航天信息研究院，中国科学院，北京 100094，中国；中国科学院大学，北京 100049，中国；海南省地球观测重点实验室，海南航空航天信息研究院，三亚 572029，中国
Xinli Hu, 中国科学院航空航天信息研究所，北京 100094，中国；中国科学院大学，北京 100049，中国；海南省地球观测重点实验室，海南航空航天信息研究所，三亚 572029，中国
Qingyan Meng, 中国科学院航空航天信息研究所，北京 100094，中国；中国科学院大学，北京 100049，中国；海南省地球观测重点实验室，海南航空航天信息研究所，三亚 572029，中国
Jiangkang Qian, 中国科学院航空航天信息研究所，北京 100094，中国；中国科学院大学，北京 100049，中国；海南省地球观测重点实验室，海南航空航天信息研究所，三亚 572029，中国
Jianfeng Gao, 中国科学院航空航天信息研究所，北京 100094，中国；中国科学院大学，北京 100049，中国；海南省地球观测重点实验室，海南航空航天信息研究所，三亚 572029，中国
Ting Li，中国空间技术研究院遥感卫星应用中心，北京市海淀区友谊路104号，邮政编码100094，中国
论文88
Corrigendum to “Biomass estimation of abandoned orange trees using UAV-SFM 3D points” [Int. J. Appl. Earth Obs. Geoinf. 130 (2024) 103931]
“使用UAV-SFM 3D点估算废弃橙树的生物量” [Int. J. Appl. Earth Obs. Geoinf. 130 (2024) 103931] 的更正
【摘要】
The authors regret that in the published article, information in CRediT authorship statement did not match that in submitted article. The correct information is J. Estornell: Conceptualization, Funding acquisition, Data curation, Methodology, Formal analysis, Investigation, Writing – original draft, Validation. J. Martí: Data curation, Methodology, Software Visualization, Supervision. E. Hadas: Methodology, Investigation, Software, Visualization, Supervision, Writing – original draft. B. Velázquez-Martí: Conceptualization, Methodology, Investigation, Writing – original draft, Supervision. I. López-Cortés: Data curation, Methodology, Supervision; A. Fernández-Sarría: Methodology, Formal analysis, Visualization, Supervision, Writing - original draft.
【摘要翻译】
作者声明: 在已发布的文章中，CRediT 作者贡献声明的信息与提交的文章不匹配。正确的信息如下：J. Estornell: 概念构思，资金获取，数据整理，方法学，正式分析，研究，原稿撰写，验证。J. Martí: 数据整理，方法学，软件，数据可视化，监督。E. Hadas: 方法学，研究，软件，数据可视化，监督，原稿撰写。B. Velázquez-Martí: 概念构思，方法学，研究，原稿撰写，监督。I. López-Cortés: 数据整理，方法学，监督。A. Fernández-Sarría: 方法学，正式分析，数据可视化，监督，原稿撰写。
【doi】
https://doi.org/10.1016/j.jag.2024.103990
【作者信息】
J. Estornell, 地理环境制图与遥感组 (CGAT)，瓦伦西亚理工大学，Camino de Vera s/n，46022 瓦伦西亚，西班牙
J. Martí, 海岸区域综合管理研究所，瓦伦西亚理工大学，Paranimf街1号，Gandia港，46730 瓦伦西亚，西班牙
E. Hadas, 弗罗茨瓦夫环境与生命科学大学，Norwida街25号，50-375 弗罗茨瓦夫，波兰
I. López-Cortés, 弗罗茨瓦夫环境与生命科学大学，Norwida街25号，50-375 弗罗茨瓦夫，波兰
B. Velázquez-Martí, 瓦伦西亚理工大学农业与食品工程系，Camino de Vera街，46022 瓦伦西亚，西班牙
A. Fernández-Sarría，地球环境制图与遥感研究组（CGAT），瓦伦西亚理工大学，Camino de Vera s/n，46022 瓦伦西亚，西班牙
